Case Information
*0 FILED IN 15th COURT OF APPEALS AUSTIN, TEXAS 4/23/2025 2:53:44 PM CHRISTOPHER A. PRINE Clerk *1 ACCEPTED 15-25-00001-CV FIFTEENTH COURT OF APPEALS AUSTIN, TEXAS 4/23/2025 2:53 PM No. 15-25-00001-CV CHRISTOPHER A. PRINE CLERK IN THE FIFTEENTH COURT OF APPEALS CREATEAI HOLDINGS, INC. (f/k/a TUSIMPLE HOLDINGS, INC.), Appellant , v.
BOT AUTO TX INC.,
Appellee . On Appeal from the Eleventh Division of the Texas Business Court Harris County, Texas
Trial Court Cause No. 24-BC11A-0007 BRIEF OF APPELLEE
Joseph A. Fischer, III, lead counsel Allison Allman Texas Bar No. 00789292 Texas Bar No. 24094023 tfischer@jw.com aallman@jw.com Leisa Talbert Peschel J ACKSON W ALKER LLP Texas Bar No. 24060414 777 Main Street, Suite 2100 lpeschel@jw.com Fort Worth, Texas 76102 Victoria C. Emery Telephone: (817) 334-7200 Texas Bar No. 24126228 Facsimile: (817) 334-7290 temery@jw.com
J ACKSON W ALKER LLP
1401 McKinney, Suite 1900
Houston, Texas 77010
Telephone: (713) 752-4530
Facsimile: (713) 308-4130
Counsel for Appellee Bot Auto TX Inc.
ORAL ARGUMENT NOT REQUESTED *2
I DENTITIES OF P ARTIES AND C OUNSEL Pursuant to T EX . R. A PP . P. 38.1(a), the following is a complete list of all parties to the trial court’s judgment and the names and addresses of all
counsel appearing in the trial or appellate courts:
Appellant
CreateAI Holdings, Inc. (formerly known as TuSimple Holdings, Inc.) Counsel for Appellant
Timothy J. McCarthy Isaac Villarreal Texas Bar No. 24123750 Texas Bar No. 24054553 tmccarthy@dykema.com ivillareal@dykema.com Cliff P. Riley Amanda Gordon Texas Bar No. 24094915 State Bar No. 24103737 criley@dykema.com agordon@dykema.com Salvador J. Robles D YKEMA G OSSETT PLLC Texas Bar No. 24121800 5 Houston Center srobles@dykema.com 1401 McKinney St., Suite 1625 Aaron J. Kotulek Houston, Texas 77010 Texas Bar No. 24137483 Telephone: (713) 904-6900 akotulek@dykema.com Facsimile: (214) 462-6401 D YKEMA G OSSETT PLLC
1717 Main Street, Suite 4200
Dallas, Texas 75201
Telephone: (214) 462-6400
Facsimile: (214) 462-6401
i
Appellee
Bot Auto TX Inc.
Counsel for Appellee
Joseph A. Fischer, III Allison Allman Texas Bar No. 00789292 aallman@jw.com tfischer@jw.com Texas Bar No. 24094023 Leisa Talbert Peschel 777 Main Street, Suite 2100 Texas Bar No. 24060414 J ACKSON W ALKER LLP lpeschel@jw.com Fort Worth, Texas 76102 Victoria C. Emery Telephone: (817) 334-7200 Texas Bar No. 24126228 Facsimile: (817) 334-7290 temery@jw.com
J ACKSON W ALKER LLP
1401 McKinney, Suite 1900
Houston, Texas 77010
Telephone: (713) 752-4530
Facsimile: (713) 308-4130
ii
T ABLE OF C ONTENTS
Identities of Parties and Counsel ................................................................... i
Table of Contents ......................................................................................... iii
Table of Authorities ..................................................................................... vii
Statement of the Case .................................................................................. xii
Statement Regarding Oral Argument ......................................................... xiv
Issue Presented [Restated] .......................................................................... xv
Statement of Facts ......................................................................................... 5
I. Factual Background ................................................................................. 5
A. TuSimple’s Beginning (n/k/a CreateAI). ........................................... 5 1. TuSimple completed a hub-to-hub demo in 2018. ...................... 6 2. Cheng Lu joined TuSimple as CFO in 2019. ................................ 7 B. TuSimple’s Turmoil: Infighting, Lawsuits, and Investigations. ........ 7 1. TuSimple explored a potential partnership with Mr. Chen’s China-based Hydron. ................................................................... 9 2. TuSimple fired Dr. Hou on October 30, 2022 and dismantled its autonomous trucking program in the summer of 2023. ........................................................................ 10 3. TuSimple’s internal “investigations” found no wrongdoing. ...... 11
C. Bot Auto’s Fresh Start: Artificial Intelligence, New Tools, and Success. ............................................................................................ 13 1. With emergent AI, Bot Auto starts fresh. .................................. 13 a. TuSimple’s convolutional neural network (CNN) had to manually train the AI model. ............................................... 14 iii
b. Bot Auto’s pretrained transformer neural network (TNN) is able to “pretrain” the AI model. .............................15
c. Bot Auto’s pretrained transformer neural network (TNN) no longer needs sensor fusion. ................................. 16 d. In addition to new AI, Bot Auto’s system and infrastructure is based on more efficient open-source technology that was not available when TuSimple’s system was created. .............................................................. 17 e. Bot Auto’s sensor technology takes advantage of new improvements in components. ............................................ 18 2. Dr. Hou leverages autonomous vehicle basics at Bot Auto. ...... 19 a. A suite of sensors gives the truck perception. ...................... 20 b. AI helps the truck make decisions. ...................................... 23 c. Redundancies make all trucks safer. ................................... 23 3. Dr. Hou protected Bot Auto’s technology from contamination by any outside source including TuSimple. .... 24 D. TuSimple’s Failure: Liquidation Dispute, Pivot to Anime, and this Lawsuit. ........................................................................................... 26 II. Procedural Background ......................................................................... 28
A. CreateAI’s Application for Temporary Injunction. .......................... 28 B. Two-Day Temporary Injunction Hearing. ....................................... 29 1. Dr. Hou testified for Bot Auto, refuting CreateAI’s speculation and claims. ............................................................. 30 2. Mr. Lu testified for CreateAI but offered no support for an injunction................................................................................... 31 3. The trial court properly denied the injunction. ......................... 33 C. Bot Auto’s Counterclaims. ............................................................... 33 iv
Summary of Argument ................................................................................ 34
Legal Standard ............................................................................................ 37
Argument .................................................................................................... 39
I. The Trial Court Properly Denied the Injunction. .................................. 39
A. CreateAI failed to state a TUTSA claim. .......................................... 39 1. CreateAI failed to identify any information entitled to trade secret protection. .............................................................. 40 a. The alleged “trade secrets” are not adequately defined. ...... 41 b. The alleged “trade secrets” are publicly available and not adequately protected. .................................................... 45 B. CreateAI failed to demonstrate a probable right to recover on its TUTSA claim. ................................................................................... 48 1. CreateAI failed to prove any alleged trade secret was acquired through improper means. ........................................... 48 2. CreateAI failed to prove any use or threatened use of an alleged trade secret. ................................................................... 50 a. CreateAI’s support— two pre-TUTSA cases examined evidence of actual trade secret possession by a competitor not present here. ................................................51 b. Bot Auto is not using, and is not in a position to use, CreateAI’s sensor -suite technology. ..................................... 56 c. Bot Auto is not using, and is not in a position to use, CreateAI’s decision -making technology............................... 57 d. Bot Auto is not using, and is not in a position to use, CreateAI’s safety technology. ............................................... 59
C. CreateAI failed to show a probable, imminent, irreparable injury. . 60 1. Injury is not probable; CreateAI’s TUTSA claim fails foundationally for lack of trade secrets. .................................... 62 v
2. Injury is not imminent (TuSimple is now CreateAI). ................ 62 3. Injury is not irreparable (CreateAI is trying to license — monetize its alleged trade secrets). ......................................... 64 D. CreateAI failed to carry its trial and appellate burdens. .................. 65 Prayer .......................................................................................................... 66
Certificate of Compliance ............................................................................ 68
Certificate of Service .................................................................................... 68
vi
T ABLE OF A UTHORITIES
Page(s) Cases
Abbott v. Anti-Defamation League Austin, Sw., & Texoma
Regions ,
610 S.W.3d 911 (Tex. 2020) (per curiam) ........................................ 37, 38 Amend v. Watson ,
333 S.W.3d 625 (Tex. App. — Dallas 2009, no pet.) .......................... 38, 59
Bianco v. Globus Med., Inc. ,
No. 2:12-CV-00147-WCB, 2014 WL 1049067
(E.D. Tex. Mar. 17, 2014) ........................................................................ 62 Butnaru v. Ford Motor Co. ,
84 S.W.3d 198 (Tex. 2002) ............................................................... 38, 65 CAE Integrated, L.L.C. v. Moov Techs., Inc. ,
44 F.4th 257 (5th Cir. 2022) ............................................................. 50, 58 Camp v. Shannon ,
348 S.W.2d 517 (Tex. 1961) .................................................................... 29 Cardinal Health Staffing Network, Inc. v. Bowen ,
106 S.W.3d 230 (Tex. App. — Houston [1st Dist.] 2003, no pet.) .................................................................................................... 55 Conley v. DSC Commc’ns Corp. ,
No. 05 – 98 – 01051 – CV, 1999 WL 89955 (Tex. App. — Dallas Feb. 24, 1999, no pet.) ............................................................................ 55 Cooper Valves, LLC v. ValvTechnologies, Inc. ,
531 S.W.3d 254 (Tex. App. — Houston [14th Dist.] 2017, no pet.) .................................................................................................... 42 Dallas Hous. Auth. v. Nelson ,
No. 05 – CV, 2015 WL 1261953 (Tex. App. Dallas Mar. 19, 2015, pet. denied) (mem. op.) .................................................. 38 vii
Davis v. Huey ,
571 S.W.2d 859 (Tex. 1978) .................................................................... 59 ECT Int’l, Inc. v. Zwerlein ,
597 N.W.2d 479 (Wis. Ct. App. 1999) ..................................................... 43 EMSL Analytical, Inc. v. Younker ,
154 S.W.3d 693 (Tex. App. — Houston [14th Dist.] 2004, no pet.) .............................................................................................. 38, 65 FMC Techs., Inc. v. Murphy ,
679 S.W.3d 788 (Tex. App. — Houston [1st Dist.] 2023, pet. denied) ............................................................................................. 46 Frey v. DeCordova Bend Estates Owners Ass ’ n ,
647 S.W.2d 246 (Tex. 1983) ................................................................... 62
Galtex Prop. Invs., Inc. v. City of Galveston ,
113 S.W.3d 922 (Tex. App. — Houston [14th Dist.] 2003, no pet.) ......................................................................................................... 64 GE Betz Inc. v. Moffitt-Johnson ,
301 F. Supp. 3d 668 (S.D. Tex. 2014) ..................................................... 42 Harbor Perfusion, Inc. v. Floyd ,
45 S.W.3d 713 (Tex. App. — Corpus Christi 2001, no pet.) .......................61
Henry v. Cox ,
520 S.W.3d 28 (Tex. 2017) ..................................................................... 38 Hilb, Rogal & Hamilton Co. v. Wurzman ,
861 S.W.2d 30 (Tex. App. — Dallas 1993, no writ) .................................. 33
IAC, Ltd. v. Bell Helicopter Textron, Inc. ,
160 S.W.3d 191 (Tex. App. — Fort Worth 2005, no pet.) ................. passim
Imax Corp. v. Cinema Techs., Inc. ,
152 F.3d 1161 (9th Cir. 1998) .................................................................. 43 INEOS Grp. Ltd. v. Chevron Phillips Chem. Co., LP ,
312 S.W.3d 843 (Tex. App. Houston [1st Dist.] 2009, no pet.) ......................................................................................................... 59 viii
Legacy Home Health Agency, Inc. v. Apex Primary Care, Inc. ,
No. 13-13-00087-CV, 2013 WL 5305238 (Tex. App. — Corpus Christi-Edinburg 2013, pet. denied) ....................................................... 62 Luccous v. J. C. Kinley Co. ,
376 S.W.2d 336 (Tex. 1964) .................................................................... 45 Mabrey v. SandStream, Inc. ,
124 S.W.3d 302 (Tex. App. — Fort Worth 2003, no pet.) ........................ 64
Metallurgical Indus. Inc. v. Fourtek, Inc. ,
790 F.2d 1195 (5th Cir. 1986) ................................................................. 43 Midway CC Venture I, LP v. O&V Venture, LLC ,
527 S.W.3d 531 (Tex. App. — Houston [1st Dist.] 2017, no pet.) ......................................................................................................... 65 N. Cypress Med. Ctr. Operating Co., Ltd. v. St. Laurent ,
296 S.W.3d 171 (Tex. App. — Houston [14th Dist.] 2009, no pet.) ..........................................................................................................61 Neurodiagnostic Consultants, LLC v. Nallia ,
No. 03-18-00609-CV, 2019 WL 4231232 (Tex. App. — Austin Sept. 6, 2019, no pet.) (mem. op.) .......................................................... 39 Nichia Corp. v. Everlight Elecs. Co. ,
No. 02:13-CV-702-JRG, 2016 WL 310142 (E.D. Tex. Jan. 25, 2016), aff’d sub nom. , Nichia Corp. v. Everlight Americas, Inc. , 855 F.3d 1328 (Fed. Cir. 2017) ................................................. 64, 65
Palliative Plus LLC v. A Assure Hospice, Inc. ,
No. 03-23-00770-CV, 2025 WL 284920 (Tex. App. — Austin Jan. 24, 2025, no pet. h.) (Field, J.) ....................................................... 39 Phazr, Inc. v. Ramakrishna ,
No. 3:19-CV-01188-X, 2020 WL 5526554 (N.D. Tex. Sept. 14, 2020) (Starr, J.) ...................................................................................... 55 Ramirez v. Ignite Holdings, Ltd. ,
No. 05-12-01024-CV, 2013 WL 4568365 (Tex. App. Dallas Aug. 26, 2013, no pet.) ............................................................................ 42 ix
Ramsey v. Sheet Pile, LLC ,
No. A-21-CV-00331-LY, 2022 WL 4000714 (W.D. Tex. Sept.
1, 2022) ................................................................................................... 44 Retail Servs. WIS Corp. v. Crossmark, Inc. ,
No. 05-20-00937-CV, 2021 WL 1747033 (Tex. App. — Dallas May 4, 2021, pet. denied) (mem. op.) ..................................................... 42 Roberson v. Robinson ,
768 S.W.2d 280 (Tex. 1989) ................................................................... 38 Shamoun & Norman, LLP v. Yarto Intern. Grp., LP ,
398 S.W.3d 272 (Tex. App. — Corpus Christi 2012, pet.
dism’d) ................................................................................................ 4, 50 Shenzhen Chengront Tech. Co., Ltd. v. Besign Direct ,
No. 1:22-CV-10281-JLR, 2022 WL 17741496 (S.D.N.Y. Dec.
9, 2022) .................................................................................................. 63 Shor v. Pelican Oil & Gas Mgmt., LLC ,
405 S.W.3d 737 (Tex. App. Houston [1st Dist.] 2013, no pet.) ......................................................................................................... 59 SL Montevideo Tech., Inc. v. Eaton Aerospace, LLC ,
No. Civ. 03-3302 RHK/FLN, 2006 WL 1472860, at *3 (D.
Minn. May 26, 2006), aff ’ d on other grounds , 491 F.3d 350 (8th Cir. 2007) ........................................................................................ 43 Spear Marketing, Inc. v. BancorpSouth Bank ,
791 F.3d 586 (5th Cir. 2015) ................................................................... 55 St. Jude Med. S.C., Inc. v. Janssen-Counotte ,
No. A-14-CA-00877-SS, 2015 WL 11438611 (W.D. Tex. Oct.
30, 2015) ............................................................................................ 55, 57 St. Jude Med. S.C., Inc. v. Janssen-Counotte ,
No. A-14-CA-877-SS, 2014 WL 7237411 (W.D. Tex. Dec. 17, 2014) ....................................................................................................... 40 State v. Hollins ,
620 S.W.3d 400 (Tex. 2020) (per curiam) ....................................... 38, 65 x
StoneEagle Servs., Inc. v. Valentine ,
No. 3:12-CV-1687-P, 2013 WL 9554563 (N.D. Tex. June 5, 2013) ....................................................................................................... 43 Sw. Rsch. Inst. v. Keraplast Techs., Ltd. ,
103 S.W.3d 478 (Tex. App. — San Antonio, 2003, no pet.) ..................... 45
T-N-T Motorsports v. Hennessey Motorsports ,
965 S.W.2d 18 (Tex. App. — Houston [1st Dist.] 1998)..................... passim
T2 Modus, LLC v. Williams-Arowolo ,
No. 4:22-CV-00263, 2023 WL 6221429 (E.D. Tex. Sept. 25, 2023) ...................................................................................................... 44 Tel. Equip. Network, Inc. v. TA/Westchase Place, Ltd. ,
80 S.W.3d 601 (Tex. App. — Houston [1st Dist.] 2002, no pet.) ......................................................................................................... 39 Tewari De-Ox Sys., Inc. v. Mountain States/Rosen, L.L.C. ,
637 F.3d 604 (5th Cir. 2011) ................................................................... 45 Topstone Commc’ns, Inc. v. Xu ,
729 F. Supp. 3d 701 (S.D. Tex. 2024) ............................................... 43, 45 Trilogy Software, Inc. v. Callidus Software, Inc. ,
143 S.W.3d 452 (Tex. App. — Austin 2004, pet. denied) ......................... 40
Walling v. Metcalfe ,
863 S.W.2d 56 (Tex. 1993) ..................................................................... 37 Wash. DC Party Shuttle, LLC v. IGuide Tours, LLC ,
406 S.W. 3d 723 (Tex. App. Houston [14th Dist.] 2013, pet. denied) ..............................................................................................61 Statutes
Tex. Civ. Prac. & Rem. Code § 134A.001 ..................................................... 39
Tex. Civ. Prac. & Rem. Code § 134A.002(6) ................................................ 40
Tex. Civ. Prac. & Rem. Code § 134A.003(a) ................................................ 44
xi
S TATEMENT OF THE C ASE
Nature of the This appeal arises from the trial court’s discretionary
Case: denial of p laintiff’s temporary injunction application
for the alleged misappropriation of trade secrets. CR.1091 – 92. [1]
Course of Dr. Xiaodi Hou (“ Dr. Hou ”) co-founded Appellant
Proceedings: TuSimple Holdings, Inc. (now known as CreateAI Holdings, Inc. ( “ CreateAI ” ) after a name change following its pivot away from autonomous trucking to anime and videogames). [2] At TuSimple, Dr. Hou developed the world’s first hub -to-hub demo for autonomous trucks. CR.0807 – 23. Amid significant interpersonal, strategic, managerial, and operational conflicts, Dr. Hou left TuSimple in 2022 and started Appellee Bot Auto TX Inc. (“ Bot Auto ”) in 2023. Bot Auto has achieved remarkable success in less time and with less funding — not because of TuSimple/CreateAI ’s alleged trade secrets (which are not secret and nonetheless outdated and useless to Bot Auto) but because Bot Auto utilizes different artificial intelligence technology, only recently made available.
On October 1, 2024, CreateAI sued Bot Auto for trade secret misappropriation under the Texas Uniform Trade Secrets Act (“ TUTSA ”) , seeking injunctive relief and damages, on the basis of its suspicion that Bot Auto’s success “could not be even colorably true” unless Bot Auto was using CreateAI’s alleged trade secrets . CR.6 – 38, 27, 50. The trial court (the Honorable Sofia Adrogué) denied CreateAI’s application for a temporary injunction on December 28, 2024. CR.1091 92.
*14 On December 30, 2024, Bot Auto filed counterclaims against CreateAI, including bad faith assertion of trade secret misappropriation, business disparagement, and tortious interference. CR.1099 – 1150. On March 17, 2025, t he trial court denied CreateAI’s Rule 91a motion to dismiss these counterclaims. SCR.592 93. The case is ongoing at the trial court level with discovery proceeding on CreateAI’s TUTSA claim and Bot Auto’s counterclaims. Bot Auto also filed a traditional and no evidence motion for summary judgment against CreateAI’s TUTSA claim based on CreateAI’s failure to identify its alleged trade secrets with reasonable particularity sufficient to distinguish the alleged “trade secrets” from common practice and publicly disclosed information, such as the disclosures in CreateAI’s extensive patent portfolio. SCR.594 – 2176. This motion is set for hearing on May 8, 2025. This appeal concerns only whether the trial court abused its discretion in denying CreateAI’s application for a temporary injunction.
Trial Court’s After written discovery, document productions, and
Disposition: depositions of each party’s principal witness (Bot
Auto’s CEO Dr. Hou and CreateAI’s CEO Cheng Lu ), the trial court (the Honorable Sofia Adrogué) resolved various discovery disputes, presided over a two-day evidentiary hearing on November 18 and 19, 2025, and considered five briefs from the parties before denying CreateAI’s application for a temporary injunction on December 28, 2024. CR.1091 92. The trial court did not issue (nor was it required to) any findings of fact or conclusions of law. CR.1094.
*15 S TATEMENT R EGARDING O RAL A RGUMENT This appeal presents no complex or novel issues of law — e.g ., trade secrets must be secret and no basis for overturning the discretionary
decision of the trial court. Appellee believes oral argument is not necessary
given the straightforward issues. Appellee welcomes the opportunity to
discuss these matters with the Court if it holds argument.
xiv *16 I SSUE P RESENTED [R ESTATED ] [4] Did the trial court abuse its discretion in denying a temporary injunction when CreateAI offered only speculation and failed to meet its
burden to plead a cause of action and show a probable right to recover with
respect to (1) trade secrets; [5] (2) Bot Auto’s use or even position to use the
alleged “trade secrets” ; [6] and (3) probable, imminent, and irreparable
injury? [7]
*17 No. 15-25-00001-CV
IN THE FIFTEENTH COURT OF APPEALS CREATEAI HOLDINGS, INC. (f/k/a TUSIMPLE HOLDINGS, INC.), Appellant , v.
BOT AUTO TX INC.,
Appellee . On Appeal from the Eleventh Division of the Texas Business Court Harris County, Texas
Trial Court Cause No. 24-BC11A-0007 BRIEF OF APPELLEE
T O THE H ONORABLE F IFTEENTH C OURT OF A PPEALS :
Laid bare, CreateAI argues, because it alleged trade secret misappropriation based on its speculation Bot Auto could not have achieved
its development milestones without stealing, CreateAI was entitled to a
presumption of harm, thus it was entitled to a temporary injunction, and
thus the trial court abused its discretion by not accepting its allegations as
true, presuming harm, and pro forma granting it a temporary injunction.
There is no right to a temporary injunction, even in a trade secrets case, nor can mere allegations or speculation support one. The evidentiary burden
to obtain a temporary injunction is high, and the burden to show there was
an abuse of discretion in denying one is even higher. CreateAI did not meet
its burden below and has not met it here.
After a two-day evidentiary hearing, in which the trial court heard testimony from Bot Auto’s Dr. Xiaodi Hou, who built its technology from the
ground up with only newly available artificial intelligence (“ AI ”) tools , and
CreateAI’s Mr. Cheng Lu, who lacks personal knowledge of the technology at
issue, the trial court determined a preliminary injunction was not warranted.
The trial court was well-within its discretion to deny CreateAI ’s application;
indeed, the evidence only supported denial.
CreateAI’s “trade secrets” are not secret; Bot Auto is not using or in a position to use them; and money damages could remedy any potential harm.
CreateAI offered only vague, highly- abstracted descriptions of its “trade
secrets” and made no effort to meet its burden to describe them with the
required detail and distinguish them from common practice and publicly
disclosed information, such as the disclosures in CreateAI’s extensive patent
portfolio. It presented no direct evidence of misappropriation, either.
In contrast, Bot Auto had no burden but presented evidence that the supposed “trade secrets” are publicly available . Bot Auto’s testimony on
public disclosure went unrebutted as did Bot Auto’s evidence it has no use
for the “trade secrets ” because its independently developed technology is
based on the latest AI that is fundamentally incompatible with CreateAI’s
now-outdated system. Bot Auto is so confident in the independent
development of its technology, it gave CreateAI access to its entire source
code for CreateAI to review and identify the allegedly misappropriated trade
secrets. 6.RR.158 59. But CreateAI never reviewed it. Id. Indeed, this “trade
secret” case presents none of the usual players— no flash drives, no customer
lists, no forensic analysis, no direct evidence at all. The usual competitive
harm is lacking too, as CreateAI has left the market and pivoted to anime and
videogames, mostly in China.
CreateAI expresses incredulousness about Bot Auto’s rapid success. But it lacks trade secrets and evidence of misappropriation; allegations and
suspicion, no matter how hyperbolic, do not support a presumption of harm
or automatic extraordinary relief. As the trial court warned from the outset,
“It’s not going to issue on mere surmise.” 2.RR.37. And yet, after weeks of
discovery, depositions, and a two-day evidentiary hearing, mere surmise is
all CreateAI offered —the trial court’s denial of the injunction was both
correct and within its discretion.
CreateAI bore the burden to present evidence of a probable trade secret claim. Instead, it speculated Bot Auto’s success “could not be even colorably
true” unless Bot Auto was using its alleged trade secrets pointing to
hyperbolic, high-level “similarities” such as sensors on the front of the
vehicle. [8] CreateAI now seeks a presumption of irreparable harm based on its
layers of speculation. [9] But harm is not presumed for a trade secret
injunction , even under CreateAI’s authorities, unless the plaintiff proves the
defendant “ possesses trade secrets ” and is “ in a position to use them .” [10] But
the alleged “trade secrets” are not secret and the evidence established Bot
Auto is not in a position to use CreateAI’s “trade secrets” as they are
platform-specific, outdated, and fundamentally incompatible with Bot
Auto’s new, AI -based approach. [11]
CreateAI failed to meet its burden to plead a cause of action and show a probable right to recover with respect to (1) trade secrets; (2) threatened
use; and (3) irreparable harm; its insistence on a presumption underscores
this. Failure to satisfy any one of those elements is fatal; CreateAI failed all
three. The trial court was well within its wide discretion to find that CreateAI
failed its burden and deny its application. Denial should be affirmed.
*21 S TATEMENT OF F ACTS
I. Factual Background
Dr. Xiaodi Hou is the world-renowned computer scientist who founded both CreateAI (where he achieved ground-breaking milestones in
autonomous trucking but whose tenure as CEO was cut short by corporate
infighting) and Bot Auto (where he is using newly available technology made
possible by recent advancements in AI). [12]
A. TuSimple’s Beginning (n/k/a CreateAI). [13] Dr. Hou graduated from Shanghai Jiao Tong University with a computer science degree, and was the first Chinese undergraduate student
to publish in multiple world-renowned AI and machine learning
conferences. 8.RR.174; 5.RR.51 – 55. Dr. Hou then joined the Computation
and Neural Systems program at the California Institute of Technology, where
he earned a Ph.D. and focused his ground-breaking research on computer
vision and machine learning, psychophysics, and electrophysiology.
8.RR.174.
*22 In 2015, Dr. Hou co-founded what is now CreateAI with Mo Chen.
8.RR.175; 5.RR.56. Dr. Hou started as Chief Technical Officer; Mr. Chen
started as CEO and Board Chairman. 8.RR.175. They evaluated different
opportunities and decided to focus on autonomous trucking in 2016. [14]
5.RR.63.
1. TuSimple completed a hub-to-hub demo in 2018. Dr. Hou and his team achieved the first hub-to-hub [15] demo for Level- 4 [16] autonomous trucks in 2018. 5.RR.63. Since then, several other
autonomous trucking companies have achieved the hub-to-hub milestone.
See 5.RR.40. But at the time, it was a remarkable feat that required an
extraordinary investment of money, time, and effort with the then-state-of-
the-art, labor-intensive technology. [17] As the first autonomous trucking
*23 company to achieve hub-to-hub, TuSimple was flooded with investments. 5.RR.225.
2. Cheng Lu joined TuSimple as CFO in 2019. In 2019, a year after the hub-to-hub milestone, Mr. Cheng Lu was hired as the Chief Financial Officer and later promoted to CEO in anticipation of
TuSimple going public. 8.RR.175; 6.RR.82. Mr. Lu used his MBA and
experience in private equity and venture capital firms to market TuSimple to
investors, which resulted in a successful IPO in April 2021. 8.RR.175;
6.RR.82.
Dr. Hou continued to develop the technology. After five years of research and development , TuSimple achieved “driver - out” operations (no
human intervention) in December 2021. 5.RR.12, 63.
B. TuSimple ’s Turmoil: Infighting, Lawsuits, and Investigations.
While Dr. Hou and Mr. Lu shared the goal of commercializing autonomous trucking, their different backgrounds, expertise, and priorities
led to an irreconcilable conflict about the best way to achieve it. 8.RR.175 –
76. Dr. Hou prioritized developing the technology to achieve profitable
commercial operations. Id. Mr. Lu focused solely on revenue , rapidly
expanding operations and offering prices well- below TuSimple’s break -even
point because he believed investors wanted startup revenue over profit. Id.
Dr. Hou viewed Mr. Lu’s revenue -over-profit strategy as short-sighted and
unsustainable. Id. He took his concerns to the Board and eventually replaced
Mr. Lu as CEO in March 2022. Id.
TuSimple’s meteoric rise as the “darling of the autonomous driving industry” was short -lived. 5.RR.225. The personality conflicts and conflicting
priorities rendered the company incapable of handling the series of setbacks
it was dealt in 2022 — including video footage of a TuSimple truck crashing
into a highway barrier, resulting in the first of many lawsuits against the
company a class action by TuSimple shareholders. See 5.RR.239 – 40, 243 –
45, 247 – 48. Then came two derivative class actions, one in early 2023 in
Delaware and one in late 2023 in California. 5.RR.240 41. All three are
ongoing. CR.1102 03.
In addition to lawsuits, the Committee on Foreign Investment in the United States (“ CFIUS ”) investigated TuSimple. 5.RR.244. CFIUS
investigates domestic entities that receive foreign investments, and
TuSimple is supported by a Chinese entity (SINA Corporation). The CFIUS
investigation resulted in TuSimple entering into a National Security
Agreement (“ NSA ”) in February 2022 to prevent proprietary information
that may implicate national security from being shared with foreign entities
(“ Covered IP ”). CR.719; 5.RR.92; 6.RR.120.
*25 1. TuSimple explored a potential partnership with Mr. Chen’s China-based Hydron.
For TuSimple to achieve commercialization, it needed to partner with an original equipment manufacturer ( “ OEM ” ). In March of 2022, when
Dr. Hou took over as CEO, TuSimple was struggling to find one. 5.RR.94 –
95. Dr. Hou met and shared information with several OEMs, including
Volvo, Hyundai, and Hydron (a Chinese company owned by TuSimple’s co -
founder, Mr. Chen). [18] Id. Dr. Hou did not disclose any Covered IP or trade
secrets to Hydron, only the information that TuSimple regularly provided to
potential OEM partners. [19]
But concern over the potential TuSimple/Hydron partnership led to two investigations one external by CFIUS into whether Covered IP was
shared in violation of the NSA and one internal led by Sullivan & Cromwell
to determine whether there was a related-party transaction. See 5.RR.227;
6.RR.120. Neither investigation found wrongdoing, but TuSimple’s capital
and reputation nonetheless were further damaged. See CR.17. Additionally,
false reports by the Wall Street Journal (claiming the FBI was investigating
TuSimple and Dr. Hou for corporate espionage) as well as rising tensions
between the U.S. and China contributed to an atmosphere of distrust at
*26 TuSimple that never improved. [20] Mr. Chen eventually left TuSimple ’s board
in June 2022 but remains an active shareholder; his investments in Chinese
anime-related entities may have influenced TuSimple’s pivot to the same
industry. 8.RR.178 – 79.
2. TuSimple fired Dr. Hou on October 30, 2022 and dismantled its autonomous trucking program in the summer of 2023.
Even though the CFIUS/Hydron investigation cleared Dr. Hou of wrongdoing, TuSimple’s Board terminated him without cause on October
30, 2022. [21]
Mr. Lu was reinstated as CEO on November 10, 2022, as TuSimple’s stock price continued to plummet from its high of $62.58 to less than $1.00
(before it eventually delisted in January 2024). 5.RR.243; CR.712. Days later,
TuSimple lost its critical manufacturing partnership with Navistar.
5.RR.243 48. TuSimple then completed the first of many reductions-in-
force, laying off hundreds of engineers responsible for its autonomous
trucking technology in December 2022. 6.RR.124.
Efforts to find a new manufacturing partner were unsuccessful, as were efforts to sell the U.S. operations, resulting in several more rounds of layoffs
*27 in the spring of 2023. 6.RR.128 – 31. Eventually, in June 2023 (more than a
year before this lawsuit was filed), TuSimple ceased all operations in the
United States, focusing on its Asia-Pacific subsidiary known as TuSimple
China. [22] 6.RR.128 – 31 (disclosing in its 8-K that it was seeking strategic
alternatives for its U.S. business). Fewer than ten of the 1,600 TuSimple
employees remain in what is left of the U.S. entity. [23]
3. TuSimple ’s internal “investigations” found no wrongdoing.
T uSimple completed two internal “investigations” after firing Dr. Hou. 5.RR.248 – 51; 6.RR.125 – 27. They found no wrongdoing and resulted in no
disciplinary action. 5.RR.248 – 51; 6.RR.125 – 27.
The first, in late 2022, involved the personal computers of TuSimple engineers who (under an old policy, when no company computers were
provided) had written TuSimple code on their personal computers. 7.RR.83 –
85; 5.RR.91. A new policy took effect in 2022 that required TuSimple work
be done on TuSimple computers. 7.RR.83 – 85. To ensure compliance with
the new policy, Mr. Lei Wang turned in his personal computer and asked
*28 TuSimple to confirm that all company information had been removed.
7.RR.83 – 85. TuSimple imaged the device, recording a full history of how,
what, and when information had been added (or removed). 5.RR.242;
7.RR.83 – 85. And Mr. Wang stayed employed at TuSimple for several
months after the uneventful “investigation , ” even receiving an accelerated
vesting of his shares as a retention bonus. 6.RR.125 27.
The second, in early 2023 after the first layoff, was initiated because a TuSimple employee reported to HR that Dr. Hou had allegedly told him,
through Mr. Wang, “that I shouldn’t be working hard … because TuSimple is
going to fail and I’m going to start a new company. ” 5.RR.249. TuSimple
hired Skadden to investigate, who interviewed a dozen employees before the
investigation was “cut short” by Dr. Hou’s resignation from the TuSimple
board. 5.RR.248 51. Other than one employee who inexplicably changed his
story in a second interview, all confirmed that Dr. Hou did not solicit or
pressure them in any way. See id. As with the laptop inquiry a few months
prior, the employee inquiry resulted in no disciplinary action or
terminations. See id.
C. Bot Auto’s Fresh Start: Artificial Intelligence, New Tools, and Success.
After being ousted by TuSimple on October 30, 2022, Dr. Hou spent the weeks after his termination “cooking and buying domain names
[including bot.auto] … and gaming. ” [24] ChatGPT was released to the public in
November 2022, weeks after his termination. 5.RR.58. And with its release,
Dr. Hou knew that AI had fundamentally changed and that he could utilize
the new technology to achieve what remained his goal, commercializing
autonomous trucking, with fewer resources and on a faster timetable. [25]
To that end, Dr. Hou decided to reenter the autonomous vehicle (“ AV ”) industry in the spring of 2023 and started operations that summer with 10 –
20 employees. 5.RR.59 61.
1. With emergent AI, Bot Auto starts fresh. Dr. Hou founded his new company with technology that did not exist when he created TuSimple’s platform. [26] Specifically, he leveraged the rapidly
advancing “transformer” technology— the same technology behind
*30 ChatGPT’s rapid success—to streamline the development of Bot Auto’s
autonomous trucking technology. [27]
a. TuSimple’s convolutional neural network (CNN) had to manually train the AI model.
T uSimple’s system was based on a “convolutional neural network” ( “ CNN ” ) architecture. 8.RR.183; 5.RR.64 –69 (“Because at that time, the new
paradigm of neural network learning ha[d] not been invented.”). The concept
of a CNN was introduced in 1989 and works by applying a mathematical
equation called “convolution” to the pixels of an image, so that it learns to
recognize patterns and relationships, eventually recognizing and predicting
images. 8.RR.183 – 84.
TuSimple’s image -recognition CNN was effective but extraordinarily expensive and time-consuming. See 8.RR.183 – 84; 5.RR.64 – 69; 6.RR.52 –
53. The unannotated data had to first be collected from video footage taken
by a mapping vehicle, and then each image had to be manually annotated
one at a time by an army of human annotators drawing a rectangle around
a “person” or a “car” until the CNN learns to recognize a “person” or a “car.” [28]
See 5.RR.67.
*31 Another limitation of TuSimple’s CNN architecture was that it only could be trained using data from the same source. 8.RR.184. For example,
the architecture that worked for 2D image processing (used for optical sensor
data) could not be utilized simultaneously for 3D data processing (used for
LiDAR sensor data). Id. As a result, TuSimple needed several different CNNs
and the time- and labor-intensive process of collecting and annotating
massive amounts of data had to be repeated for each one. Id. TuSimple’s
CNNs also required “sensor fusion” to translate the different streams of
annotated data from each network. 5.RR.75 – 76. So, while groundbreaking
at the time it was created almost seven years ago, TuSimple’s CNN s required
an extraordinary amount of time and money to build and maintain —
extending the path to and cost of commercialization. 8.RR.183 – 84;
5.RR.64 69; 6.RR.52 53.
b. Bot Auto’s pretrained transformer neural network (TNN) is able to “pretrain” the AI model.
The “transformer” architecture utilized by Bot Auto was made available to the public in late 2022, [29] after TuSimple terminated Dr. Hou, and years
after TuSimple created its CNN architecture. [30] Instead of manually training
*32 the AI model to recognize a bicycle by circling the bicycles from collected
images (as under a CNN), the transformer neural network ( “ TNN ” ) trains
itself, with very little human assistance, using images available on the
internet by covering up one image of a sequence and “rewarding” the model
for correct guesses and “punishing” for incorrect guesses until it is able to
recognize images without assistance. See 5.RR.69 – 74. The manual,
expensive, time-consuming training under TuSimple’s CNN s could not be
more different from the automated, efficient, and cost-effective pre-training
under Bot Auto’s TNN . 5.RR.190 91.
c. Bot Auto’s pretrained transformer neural network (TNN) no longer needs sensor fusion.
Bot Auto’s TNN is fundamentally different from TuSimple’s CNN because it eliminates the need for extensive manual training. It also
eliminates the need for sensor fusion because it no longer needs multiple
networks to process different kinds of data. 5.RR.76 (“the very infamous
topic called sensor fusion does not even exist if you’re using a transformer -
based neural network. You do not need to have that.”). In other words, to a
[TuSimple] vehicle”), 187–88 (the “foundation of Bot Auto’s new technology” did not exist
when he built TuSimple ’s ). Transformer pretraining started in the natural language processing domain.
5.RR.69 74. For example, ChatGPT trains its model by having it “read” the limitless
material available on the internet, such as NYT articles, covering up certain words and
rewarding correct connections. Id.
transformer, “[e]very input is a[n] input.” Id. The new TNN is like a much-
needed translator so while the CNN has to manually translate, word-by-
word, ten different languages (or “networks” of data) , one at a time, the TNN
speaks every language so every stream of data can be processed faster,
cheaper, and with more accuracy. See 5.RR.75 – 77; see also 8.RR.184 – 87.
The divergent data formats mean that TuSimple’s technology “has zero value
to Bot Auto.”
d. In addition to new AI, Bot Auto’s system and infrastructure is based on more efficient open-source technology that was not available when TuSimple’s system was created.
Further, Dr. Hou has also taken advantage of recent improvements in the AI ecosystem, specifically the open-source community. 8.RR.187;
5.RR.83 85, 191. At TuSimple, Dr. Hou employed 100 – 200 employees for
more than a year to build and maintain its software infrastructure platform
in-house. 5.RR.83 85. At Bot Auto, he has embraced open-source products
like Kubernetes and Infra as Code (“ IaC ”) to build Bot Auto’s platform with
only eight employees in four months. Id. In other words, Bot Auto is utilizing
*34 different coding technology, methods, and tools to work towards
commercialization on a faster timetable with fewer resources. 6.RR.52 – 53;
8.RR.183 – 87 (the open-source platform allows Bot Auto to handle larger
data workloads by automating its data computing, allowing the development
environment to run on a self-maintained data center and cloud storage);
5.RR.64 77, 85 ( “open source solutions were not even available [] when I
found TuSimple [] GPU computing on the cloud was not a thing back then.” ).
This fundamental change in the AI ecosystem and how Bot Auto is developing and housing its technology ( i.e. , open-source instead of in-house)
is another reason TuSimple’s technology would be useless to Bot Auto.
5.RR.83 85, 90, 191 (“Even if I got a direct hard copy of TuSimple’s source
code in my hand, it’s useless to Bot Auto.”).
e. Bot Auto’s sensor technology takes advantage of new improvements in components.
Finally, Bot Auto also leverages new sensor technology that enables it to “see further” than TuSimple. 5.RR.106. For example, Bot Auto’s high -
resolution LiDARs have over three times the resolution of TuSimple’s low -
resolution LiDARs. Id. ; 8.RR.187. Bot Auto’s more advanced sensors, data,
and AI algorithms mean that data is collected and stored in completely
different ways at the two companies. See 5.RR.106.
But while many aspects of AV technology are constantly changing, [33] some things remain the same, like the basic components both TuSimple and
Bot Auto (and every other AV company) embrace so it is no surprise both
companies use sensors and mount them on racks pointed in the direction of
travel, for example. [34]
2. Dr. Hou leverages autonomous vehicle basics at Bot Auto.
CreateAI has alleged three buckets of “trade secrets , ” defining them only at a high-level with general terms that encompass the core
characteristics of any autonomous vehicle ( “ AV ” ) ( e.g. , sensors pointed
forward). CR.7.
All AVs require certain fundamental components. 8.RR.179 80. Unsurprisingly, these include sensors, such as optical (cameras), LiDAR, and
radar, which gather data about the vehicle’s surroundings, enabling it to
detect obstacles, pedestrians, and other vehicles. Id. (“TuSimple Sensor Suite
Technology”). Additionally, computer systems process this data in real-time,
allowing the vehicle to make informed decisions and navigate complex,
*36 dynamic environments. CR.7. (“TuSimple Autonomous Decision-Making
Technology”). Control systems are also crucial to every AV, as they translate
processed information into actions (steering, braking, acceleration, etc.);
and redundant systems ensure safety when a critical component fails. Id.
(“TuSimple Automatic Safety Technology”). Together these core
components , which TuSimple has identified as its alleged “trade secrets,”
form the backbone of any AV technology. Id.
a. A suite of sensors gives the truck perception. One obvious requirement for any AV is that it be capable of “seeing” around the entire vehicle. Just as human drivers are instructed to turn their
heads to check blind spots, AV sensors must be placed to ensure 360-degree
coverage. The sensor arrangement will vary depending on two variables:
(1) the individual sensor used (all have different fields of view and
resolutions) and (2) the design of the vehicle on which they are attached.
8.RR.179. Bot Auto uses different sensors than TuSimple (as explained supra
I.C.1.e.) and different vehicles (Daimler and Freightliner instead of Peterbilt
and International Truck/Navistar). See 5.RR.100 01; 6.RR.123. So while Bot
Auto uses a suite of sensors (as all AVs do), the selection and arrangement of
those sensors differs in every way from TuSimple’s , as does the processing of
the information. 5.RR.100 01. *37 Importantly, AV manufacturers do not make their own sensors.
Instead, they buy them off-the-shelf . CR.577 – 80; 6.RR.98 – 99. With
variations in sensor-selection and sensor-placement, there are as many
different “sensor - suite” arrangements as there are vehicle manufacturers
and the images below depict some of them. CR.528 31; 582 99.
As shown in the figures above and explained by Dr. Hou, all autonomous trucks have a front-facing camera array housing on the top roof.
8.RR.179 80. The rear-view cameras are either embedded in the main
housing (Waymo, Torc, and Bot Auto) or become separate units affixed on
the body (Aurora and TuSimple). Id. All trucks have at least two “hockey
puck” shaped LiDAR units, either installed at the roof (Waymo, Aurora, Torc,
and Bot Auto), or at the engine hood (TuSimple). Id. And for safety, near-
range blindspot LiDARs are installed in some vehicles, either at the front of
the engine hood (Waymo), or the rear of the engine hood (Aurora and Bot
Auto), or at the back of the vehicle body (TuSimple). Id.
The sensor-suite design — choosing the technical specifications and installation location of each sensor is the first step in developing a sensor
algorithm. Id. It also determines the fail-safe policy (what redundancy to use
when one sensor fails). Id. And as the above figures demonstrate, Bot Auto
and TuSimple utilize different sensors and arrange them differently on
different truck brands. Id.
*39 b. AI helps the truck make decisions.
In addition to a sensor-suite, all AVs have software called neural networks that process data from the different sensors, allowing it to make
real-time decision s and avoid hazardous conditions the industry calls “edge”
or “corner” cases. 8.RR.182. And, as explained above, Bot Auto’s TNN is
fundamentally different from the CNN used at TuSimple. See supra I.C.1.
For example, “recognizing a person on a wheelchair,” or “interacting with tumble weeds” are classic examples of an “edge” or “corner” case in the
AV industry. 8.RR.182. Because each company’s neural networks are
different and every sensor has unique characteristics, the solutions are not
one-size-fits-all. Id. Different AVs have varying abilities to solve any
particular problem. Id. For example, one system might have difficulty
detecting (and thus avoiding) a small animal on the side of the road because
its sensors are arranged in a way that makes detecting low-to-the-ground
objects more challenging. Id. As a result, the solution for that edge case will
be different for each system. Id.
c. Redundancies make all trucks safer. Finally, in addition to “sensing” obstacles and “deciding” how to avoid them, AVs require redundancies to ensure safety in the event of a component
failure. [36] Redundancy is a basic and ancient rule of engineering. [37] The more
redundancies that are built into a system, the safer it will be. CR.532. And
AV technology is no different. [38]
But as with the sensor and decision-making technologies, the redundancies for every AV company are fundamentally different from each
other. 5.RR.112 – 13. For example, TuSimple and Bot Auto use different
truck brands and different brands of steering and braking, so all the
interfacing protocols and safety failure modes ( i.e. , redundancies) are
different. See id. (Dr. Hou explaining both companies have redundancies but
they use a completely different vocabulary and application to determine and
solve component-failures, similar to the safety systems of a Camry versus an
Accord — both have redundancies, but they are unrecognizable to each other).
3. Dr. Hou protected Bot Auto’s technology from contamination by any outside source including TuSimple.
In building Bot Auto’s technology from the ground -up with the newest advancements in AI, Dr. Hou has made every effort to protect it from
contamination or misappropriation by any outside source, including
*41 TuSimple. 5.RR.87 – 91, 175, 179. To be clear, Dr. Hou put in place procedures
and policies to ensure that no TuSimple trade secrets would be used at Bot
Auto. [39] 5.RR.87 – 91, 175, 179.
First, in hiring engineers, Dr. Hou was careful not to solicit those who were currently working at TuSimple. 5.RR.80 83. But in the more than
1,000 employees who have left or been let go from TuSimple since December
2022, it is no surprise that some of them have landed at Bot Auto. [40] Id. The
30 40 former TuSimple employees who now work at Bot Auto were not
solicited by Dr. Hou but either (1) were fired and facing deportation when
their short H-1B visa grace period expired; (2) voluntarily contacted
Dr. Hou; or (3) left TuSimple, worked at a different company, and eventually
came to Bot Auto. Id.
Second, Dr. Hou ensured that no employee brought trade secrets from a prior employer, including TuSimple, with them to Bot Auto. To that end,
Bot Auto has its employees sign a “Mandatory Pre -Onboarding
Questionnaire and Certification” and “Onboarding Questionnaire and
Certification” “to ensure that no confidential or proprietary information is
*42 used at Bot Auto.” 7.RR.68 – 76; 5.RR.88 – 89.
And finally, even within Bot Auto, Dr. Hou has put in place security measures to protect the new technology from contamination by any outside
source. 5.RR.87 – 91, 175, 179. For example, Bot Auto ’s system requires
manual, internal inputs and doesn’t allow additions from any outside source.
5.RR.173, 179, 90 (“if you just plug in a USB hard drive, it won’t work”). Bot
Auto ’s source code is constantly monitored and audited, built to trigger an
alarm if anyone attempts to upload or download data in bulk. 5.RR.91, 175,
179, 181.
D. TuSimple’s Failure: Liquidation Dispute, Pivot to Anime, and this Lawsuit.
By the summer of 2023, TuSimple’s path to commercialization was no longer viable. See 6.RR.128 – 31. It had voluntarily delisted its shares, shut
down all U.S. operations, and pivoted to its Asia-Pacific subsidiary, TuSimple
China. Id. ; see also 5.RR.244; 6.RR.116, 122, 131. Many shareholders
(including Dr. Hou) wanted to liquidate the company and distribute what
little value remained. 8.RR.178 – 79. This led to more disputes and the
largest shareholder, Mr. Chen, blocked dissolution. See id. ; CR.541 – 42.
Between the summer of 2023 and summer of 2024, TuSimple began to quietly pivot to a new industry familiar to Mr. Chen from other investments
(anime and videogames). See CR.541 42; see also 8.RR.178 79. During this
time, TuSimple “heard rumors” that Dr. Hou was starting a new autonomous
trucking company. See 5.RR.253. But it expressed no concern and took no
action until it received an email from a shareholder with ties to Mr. Chen
asking it to “investigat[e]” their “strong[] susp[icion s ]” that Dr. Hou was
utilizing TuSimple technology. 8.RR.15. The August 12, 2024 email offered
the shareholder had received a “teaser” from Bot Auto , (7.RR.316), and that
based on the teaser they “ strongly suspect[ed] ” that Dr. Hou was using
TuSimple’s AV technology. Id. The letter requested that TuSimple “initiate a
commercial investigation … and, if necessary, file a lawsuit to provide
shareholders with a fair and reasonable explanation.” Id.
Mr. Lu hired an investigator in August 2024 who took pictures of Bot Auto ’s trucks and confirmed several innocuous facts: Bot Auto had
operations in Houston and was operating a sensor-equipped vehicle on
public roads. See 6.RR.122; 8.RR.170 – 72.
At the same time, TuSimple accelerated its pivot from autonomous trucking to anime and videogames. See SCR.32 37. Nonetheless, it filed this
lawsuit (as TuSimple). And its rebrand, along with the name change to
CreateAI, was not made public until weeks after the temporary injunction
hearing, in December 2024. SCR.21 27.
According to CreateAI’s press release, it entered “into the digital entertainment content industry in August 2024 ” (two months before this
lawsuit was filed). Id. CreateAI now focuses “on transforming animation and
video game industries through proprietary AI technologies and its integrated
content development platform.” Id. Its new mission statement and press
release detail accomplishments with several open-world role-playing games
and anime designers but make no mention of autonomous driving. See id. ;
see also SCR.32 37. The only vestige of CreateAI’s beginning as an AV
company is found in its (so far unsuccessful) efforts to “partner” with other s
in licensing its AV technology. SCR.51; 6.RR.77.
II. Procedural Background
A. CreateAI ’s Application for Temporary Injunction. CreateAI filed this suit on October 1, 2024 (nearly two years after Dr. Hou’s departure). CR.6 172. The initial hearing was held the next day, in
which CreateAI ’s counsel previewed its now -shuttered AV business and the
reparable nature of any potential injury as the company “at this stage of its
life and its development, is largely in the business of licensing its
technologies.” 2.RR.42. The trial court warned CreateAI that an injunction
was “not going to issue on mere surmise” and noted that whether the
*45 claimed material was truly a trade secret would be “something that is worthy
of discussion.” 2.RR.59. The trial court also struggled “to appreciate the true
urgency ” … “[i]t’s just unclear to me why this has to happen now.” 2.RR.3 8 –
39 (“what would be different today versus from 14 days from now”) .
The parties agreed to a temporary restraining order that would preserve the status quo until an evidentiary hearing could be held and allow
discovery by agreement, [42] engaged in that discovery, and filed five rounds of
briefing on the application, both before and after the temporary injunction
hearing. [43] Indeed, unlike most temporary injunction hearings that are held
shortly after the application is filed with no discovery, CreateAI had the
benefit of written discovery, Bot Auto’s production of thousands of
documents, and Dr. Hou’s deposition.
B. Two-Day Temporary Injunction Hearing.
On November 18 and 19, 2024, the trial court held a two-day evidentiary hearing on CreateAI ’s application. 5.RR.4 – 289; 6.RR.4 – 176. The
CEOs of both parties Dr. Hou and Mr. Lu — testified at the hearing. [44]
*46 1. Dr. Hou testified for Bot Auto , refuting CreateAI’s speculation and claims.
Dr. Hou gave extensive testimony on his expertise in AI models, explaining the monumental, truly transformative changes the industry has
undergone in the last two years that made it possible for him to
independently develop Bot Auto’s technology from the ground-up.
5.RR.51 – 121, 77. Dr. Hou explained the fundamental differences between the
two platforms (both of which he built), why they are incompatible with each
other, and why any CreateAI “trade secrets” would be useless at Bot Auto.
5.RR.64 – 90, 187 – 88, 190 – 91.
Dr. Hou also explained how the recent advancements allowed him to build a platform that is more accurate than CreateAI ’s and can be developed
on a shorter timetable with fewer resources making commercialization
possible. 5.RR.69 – 74. He also explained how CreateAI ’s “trade secrets” are
defined so broadly that they merely describe the fundamental components
of every AV. 5.RR.96 (explaining all AVs have sensors just as they all have
wheels). Dr. Hou consistently testified throughout the hearing that Bot Auto
has never used and would have no use for any CreateAI trade secrets.
*47 2. Mr. Lu testified for CreateAI but offered no support for an injunction.
Mr. Lu was unable to refute Dr. Hou’s testimony . He does not even
have personal knowledge of or experience with the technology of either
company. [46] He joined CreateAI as the CFO months after Dr. Hou developed
the technology at issue and has never typed a single line of CreateAI source
code. 6.RR.82, 120. Mr. Lu was the only witness CreateAI offered at the
hearing and he admitted to having “no evidence” of any document that was
taken from CreateAI to Bot Auto. 6.RR.125. His testimony on CreateAI ’s
technology was as high-level as you would expect from a finance executive
and did not refute any of Dr. Hou’s testimony on the fundamental
incompatibility of the two companies ’ different neural networks. [47]
Seeking to rebut Dr. Hou’s testimony that no transformer technology
was used on a CreateAI (TuSimple) truck, Mr. Lu tried to create confusion
Q: What about the autonomous driving decision-making at Bot Auto, is that based on
TuSimple’s trade secrets? A: No, not at all. … Q: … is Bot Auto’s safety technology … based
on TuSimple’s trade secrets ? A: Not at all. ” ); 5.RR.190 ( “ Q: Even if Bot Auto had
TuSimple’s trade secrets, could Bot Auto use those trade secrets? Would they be useful to
Bot Auto? A: No … it has zero value to Bot Auto.”). 6.RR.83, 97 98 (Mr. Lu admitting he does not have the same level of expertise in
AI as Dr. Hou and he has no knowledge of what sensors Bot Auto uses, no knowledge of
whether Bot Auto’s rearview cameras are in the same place as CreateAI’s , and no
knowledge of how Bot Auto implements its cameras or sensors). See, e.g. , 6.RR.25 (noting that the Bot Auto van at 8.RR.126 “looks identical to TuSimple
… it’s likely that their camera choices, maybe not the exact model, but their types of
cameras replicates our innovations. Because TuSimple our innovation is we’re camera -
centric. ); 6.RR.27 (“[t] hey use quite a bit of cameras, which are TuSimple innovations”).
over whether transformer technology was explored at CreateAI. 5.RR.60 – 61,
95 (referring to bird’s eye view technology) . But when asked directly, Mr. Lu
acknowledged it was CreateAI’s China entity that explored using the
technology and that it did not migrate to the U.S. entity until 2023, months
after Dr. Hou was fired (a timeframe that also is consistent, as it turns out,
with the availability of this technology in the market generally, e.g. , the
emergence of ChatGPT). [48] 6.RR.94 – 96. The technology TuSimple
(CreateAI) may have been exploring through its separate, Asia-Pacific entity,
was never put on a TuSimple (CreateAI) truck in the United States and
certainly not when Dr. Hou was there. 5.RR.161 – 62; see 6.RR. 94 – 96, 106 –
07; see also 6.RR.166. [49]
Mr. Lu also acknowledged that CreateAI has “one of the largest organically developed patent portfolios in AV trucking” with more than 200
issued patents and 400 pending. 6.RR.107, 111. And he admitted that
CreateAI made no effort whatsoever to determine what part, if any , of its
alleged “trade secrets” are or are not subject to one of its hundreds of public
patents. 6.RR.111 – 16.
*49 3. The trial court properly denied the injunction.
On December 28, 2024, the trial court denied CreateAI ’s application for a temporary injunction by written order. CR.1091 – 92. It was not required
to, nor did it, issue findings of fact or conclusions of law.
C. Bot Auto ’s Counterclaims.
Also in December, Bot Auto filed counterclaims against CreateAI, including bad faith assertion of trade secret misappropriation, business
disparagement, and tortious interference. CR.1099 – 1150. CreateAI moved to
dismiss the counterclaims under Rule 91a, and, conspicuously in
contradiction of any imminent or irreparable injury, “request[ed] a stay of
the action” [50] to permit an interlocutory appeal if its motion were denied. [51]
SCR.587. On March 17, 2025, the trial court denied CreateAI ’s motion to
dismiss and its corresponding request for a stay. SCR.592 93. Accordingly,
CreateAI ’s TUTSA claim (now subject to a motion for summary judgment for
failure to distinguish the alleged “trade secrets” from common practice and
publicly disclosed information) and Bot Auto’s counterclaims continue to
proceed before the trial court. SCR.594 2176.
*50 S UMMARY OF A RGUMENT
As CreateAI ’s authorities confirm, it had the burden to plead a cause of action under TUTSA and show a probable right to recover with evidence
tending to sustain it and show a probable, imminent, and irreparable injury
in the interim. [52] This required it satisfy three distinct elements, with
evidence , and it failed on all three. CreateAI did not state a TUTSA claim
because it failed to identify any actionable trade secrets; [53] did not prove a
probable TUTSA claim because it failed to show Bot Auto is in a position to
use any “trade secrets” ; [54] and failed to prove any probable threat of
imminent, irreparable harm. [55]
Further to its own lack of evidence, CreateAI also failed to rebut Bot Auto’s direct evidence— from the founder of both companies and creator of
the technologies at issue —that the “trade secrets” are public and that, even if
some portion were kept secret (of which there is no evidence), Bot Auto has
no access to that unidentified “secret” portion and it would nonetheless have
*51 no use for such “secrets” because its cutting-edge technology is
fundamentally incompatible with CreateAI ’s , which is outdated in any event.
In a case over the alleged misappropriation of autonomous trucking “trade secrets,” CreateAI offered only one witness, Mr. Lu, who was unable
to describe the technology of either party. Mr. Lu has no personal knowledge
of CreateAI ’s outdated technology (he joined after it was built and does not
even have the ability to view the source code), and he has no personal
knowledge of Bot Auto’s cutting -edge technology. CreateAI could not, and
did not, refute Dr. Hou’s testimony that Bot Auto does not have, never h as
had, and would have no use for, the alleged “trade secrets” of CreateAI.
Mr. Lu ’s testimony focused— not on evidence of misappropriation but on frustration with the “fresh start” Dr. Hou got at Bot Auto. 5.RR.283 –
86. But that fresh start evidences Bot Auto’s independently developed
technology with the latest advancements in AI, not harm to or
misappropriation of CreateAI’s outdated “trade secrets . ”
When asked directly “Why has TuSimple brought this lawsuit?” Mr. Lu
explained that CreateAI has spent “approximately $1.4 billion, to develop
technology for self-driving. We’ve accumulated one of the largest organic
patent portfolios . We have assets, know-how, that you know, source codes,
tool chain, data that is valuable. And we want to monetize that value .”
5.RR.288 (emphasis added). But, here, there is no trade secret
misappropriation or irreparable injury to halt, no monetization to be had.
This explains why CreateAI’s arguments are based entirely on its own speculation. On speculation alone, CreateAI urges this Court to reverse the
trial court’s discretion . Instead, Texas law requires every party seeking an
injunction to meet its extraordinary burden with evidence. And CreateAI
failed to do so.
There is no evidence of trade secrets (to the contrary — the vague definitions offered by CreateAI include components that are physically
visible to the public, subject to its own public patents, or fundamental to any
autonomous vehicle). There is no evidence of threatened use (to the
contrary —CreateAI’s outdated technology is useless to Bot Auto). And there
is no threat of irreparable harm (to the contrary — CreateAI is not competing
with Bot Auto because it is not pursuing autonomous truck development, has
pivoted to anime and videogames in China, and is now only trying to license
its “trade secrets” to third parties).
CreateAI’s insistence on a presumption underscores its lack of evidence to meet its burden. Citing IAC, Ltd. v. Bell Helicopter Textron, Inc. ,
160 S.W.3d 191 (Tex. App. Fort Worth 2005, no pet.), CreateAI argues it
*53 was entitled to a presumption of harm and irreparable injury. [57] But, Bell
Helicopter is clear a presumption of harm is available only when “a
defendant possesses trade secrets and is in a position to use them .” [58] And
the evidence, unaddressed by CreateAI, established the alleged “trade
secrets” are not secret and Bot Auto is not in a position to use them in any
event as it is using different, new, AI-based technology that is fundamentally
different from and incompatible with TuSimple’s outdated technology.
The trial court properly exercised its discretion in denying the injunction. Any other result would have been reversible error. The order
should be affirmed because CreateAI did not carry its burden for the
extraordinary relief it sought.
L EGAL S TANDARD “ A temporary injunction is an extraordinary remedy and does not issue as a matter of right. ” Abbott v. Anti-Defamation League Austin, Sw., &
Texoma Regions , 610 S.W.3d 911, 916 (Tex. 2020) (per curiam) (quoting
Walling v. Metcalfe , 863 S.W.2d 56, 57 (Tex. 1993)). The party applying for
a temporary injunction bears the burden to “ prove three specific elements:
(1) a cause of action against the defendant; (2) a probable right to the relief
sought; and (3) a probable, imminent, and irreparable injury in the interim. ”
*54 State v. Hollins , 620 S.W.3d 400, 405 (Tex. 2020) (per curiam) (quoting
Butnaru v. Ford Motor Co. , 84 S.W.3d 198, 204 (Tex. 2002)). A probable
right of recovery is shown by alleging a cause of action and presenting
evidence tending to sustain it. Bell Helicopter , 160 S.W.3d. at 200.
A trial court’s order denying a temporary injunction is reviewed only for abuse of discretion. Hollins , 620 S.W.3d at 405. And in resolving
evidentiary matters, “a trial court does not abuse its discretion ‘ if some
evidence reasonably supports the court’s ruling . ’” Abbott , 610 S.W.3d at 916
(quoting Henry v. Cox , 520 S.W.3d 28, 34 (Tex. 2017)). Abuse-of-discretion
review requires that the appellate court “view the evidence in the light most
favorable to the trial court’s order, indulging every reasonable inference in
its favor,” and defer to the trial court’s resolution of conflicting evidence.
Amend v. Watson , 333 S.W.3d 625, 627 (Tex. App. — Dallas 2009, no pet.).
When, as here, there are no findings of fact or conclusions of law, the appellate court implies all necessary findings in support of the trial court’s
decision and upholds it on any legal theory supported by the record. EMSL
Analytical, Inc. v. Younker , 154 S.W.3d 693, 696 (Tex. App. — Houston [14th
Dist.] 2004, no pet.); Dallas Hous. Auth. v. Nelson , No. 05 – CV,
2015 WL 1261953, at *2 (Tex. App. Dallas Mar. 19, 2015, pet. denied) (mem.
op.) (citing Roberson v. Robinson , 768 S.W.2d 280, 281 (Tex. 1989)).
A RGUMENT
I. The Trial Court Properly Denied the Injunction.
A failure on any required element supports the trial court’s discretionary denial; CreateAI failed across the board.
A. CreateAI failed to state a TUTSA claim.
CreateAI ’s claim is for misappropriation of trade secrets under TUTSA. CR.29 31; Tex. Civ. Prac. & Rem. Code §§ 134A.001 008. To establish a
probable right to recover on this claim, CreateAI bore the burden to present
evidence on each element. See Tel. Equip. Network, Inc. v. TA/Westchase
Place, Ltd. , 80 S.W.3d 601, 607 (Tex. App. — Houston [1st Dist.] 2002, no
pet.); Bell Helicopter , 160 S.W.3d at 200.
The elements of a TUTSA claim are:
(1) the existence of a trade secret;
(2) acquisition of the trade secret through a confidential relationship or discovery of it by improper means; (3) use of the trade secret without authorization; and (4) resulting damages to the plaintiff.
Palliative Plus LLC v. A Assure Hospice, Inc. , No. 03-23-00770-CV, 2025
WL 284920, at *11 (Tex. App. — Austin Jan. 24, 2025, no pet. h.) (Field, J.)
(citing Neurodiagnostic Consultants, LLC v. Nallia , No. 03-18-00609-CV,
2019 WL 4231232, at *8 (Tex. App. Austin Sept. 6, 2019, no pet.) (mem.
op.) (citing Trilogy Software, Inc. v. Callidus Software, Inc. , 143 S.W.3d
452, 463 (Tex. App. — Austin 2004, pet. denied))).
1. CreateAI failed to identify any information entitled to trade secret protection.
Under TUTSA, information qualifies as a “trade secret” only if it meets two requirements:
(A) [T]he owner of the trade secret has taken reasonable measures under the circumstances to keep the information secret ; and
(B) the information derives independent economic value, actual or potential, from not being generally known to, and not being readily ascertainable through proper means by, another person who can obtain economic value from the disclosure or use of the information.
Tex. Civ. Prac. & Rem. Code § 134A.002(6) (emphasis added).
While an applicant is not required to prove the alleged trade secrets are in fact trade secrets until trial, it still bears the burden at the temporary
injunction stage to identify information that more likely than not satisfies
TUTSA i.e. , merits trade secret protection pending a trial on the merits.
The first step of doing so requires a plaintiff to identify with reasonable particularity the metes and bounds of information that allegedly warrants
trade secret protection. St. Jude Med. S.C., Inc. v. Janssen-Counotte , No. A-
14-CA-877-SS, 2014 WL 7237411, at *15 (W.D. Tex. Dec. 17, 2014). CreateAI
did not and cannot meet this basic burden to identify its “trade secrets”— the
basis of Bot Auto’s pending motion for summary judgment. SCR.594 – 2176.
a. The alleged “trade secrets” are not adequately defined.
CreateAI’s claim must start with adequately defining its alleged “trade secrets” ; its failure to do so is dispositive. Indeed, the trial court keyed in
on this issue at the initial hearing. 2.RR.59 (noting that whether the
application described true trade secrets would be “something that is worthy
of discussion.”). But weeks later, after written discovery, depositions, and a
two-day evidentiary hearing, CreateAI still had not adequately defined its
“trade secrets . ”
Adequately defining the “trade secrets” at issue is an essential first step of both stating a TUTSA claim and seeking an injunction because an
injunction must specifically describe the alleged trade secrets and give
*58 adequate notice of the restrained acts. Ramirez v. Ignite Holdings, Ltd. , No.
05-12-01024-CV, 2013 WL 4568365, at *4 (Tex. App. — Dallas Aug. 26, 2013,
no pet.); Cooper Valves, LLC v. ValvTechnologies, Inc. , 531 S.W.3d 254,
265 66 (Tex. App. — Houston [14th Dist.] 2017, no pet.).
Courts frequently refuse to issue or overturn injunctions when the plaintiff fails to adequately define the trade secrets at issue, as CreateAI has
failed to do here. Retail Services WIS Corp. v. Crossmark, Inc. , No. 05-
20-00937-CV, 2021 WL 1747033, at *14 (Tex. App. Dallas May 4, 2021, pet.
denied) (mem. op.) (reversing portions of an injunction when the trade
secret definition was “too vague”); see also Ramirez, 2013 WL 4568365, at
*4 (vacating portions that did “not define ‘Proprietary Information/Tr ade
Secrets’ with enough specificity to give appellants adequate notice of the acts
they are restrained from doing.”).
The broad, abstract definitions offered by CreateAI of its alleged trade secrets —“Sensor Suite Technology,” “Autonomous Decision -Making
Technology,” and “Automatic Safety Technology”— generically encompass
the macro components of every AV on the road today ( i.e. , ubiquitous,
general knowledge in the AV industry). GE Betz Inc. v. Moffitt-Johnson ,
301 F. Supp. 3d 668, 691 (S.D. Tex. 2014) (“[M]atters of general knowledge
*59 in an industry cannot be appropriated by one as his secret.”) (citing
Metallurgical Indus. Inc. v. Fourtek, Inc. , 790 F.2d 1195, 1199 (5th Cir.
1986)); Topstone Commc’ns, Inc. v. Xu , 729 F. Supp. 3d 701, 706 07 (S.D.
Tex. 2024) (dismissing trade secret claim because the allegations did not
separate plaintiff’s information “from matters of general knowledge”) .
CreateAI’s generic, high-level, kitchen-sink approach to defining the alleged “trade secrets” falls far short of its burden to describe the precise
information at issue with reasonable particularity. See StoneEagle Servs.,
Inc. v. Valentine , No. 3:12-CV-1687-P, 2013 WL 9554563, *2 *5 (N.D. Tex.
June 5, 2013) (recognizing that Texas law requires plaintiffs “to identify, with
reasonable particularity, the alleged trade secrets at issue” and explaining “a
list that separately breaks out each of the individual alleged trade secrets and
identifies each claim so that the reader understands how each claim differs
from public domain filings” complies, but generalized disclosures that only
*60 reveal “the end results of, or functions performed by, the claimed trade
secrets” do not); T2 Modus, LLC v. Williams-Arowolo , No. 4:22-CV-00263,
2023 WL 6221429, *3 – *5 (E.D. Tex. Sept. 25, 2023) (same).
CreateAI’s vaguely -defined and highly-abstracted “trade secrets” make its true aim clear: stunt Bot Auto’s success by broadly enjoining the “know -
how” they complained of at the temporary injunction hearing i.e. , general
knowledge and experience that TUTSA makes clear cannot be enjoined. [62]
Ramsey v. Sheet Pile, LLC , No. A-21-CV-00331-LY, 2022 WL 4000714, at *5
(W.D. Tex. Sept. 1, 2022) (“Texas law restricts injunctions from
‘prohibit[ing] a person from using general knowledge, skill, and experience
that person acquired during employment.’”) (quoting Tex. Civ. Prac. & Rem.
Code § 134A.003(a)). Dr. Hou’s “know - how” is not a trade secret as a matter
of law and the trial court properly refused to enjoin that general knowledge
and experience. See id. ; 6.RR.166.
*61 b. The alleged “trade secrets” are publicly available and not adequately protected.
In addition to lacking specificity, the “trade secrets” claimed by CreateAI also lack the essential element of secrecy — another dispositive
failure. Tewari De-Ox Sys., Inc. v. Mountain States/Rosen, L.L.C. , 637 F.3d
604, 611 (5th Cir. 2011) (“It is self -evident that the subject matter of a trade
secret must be secret.”) (quoting Luccous v. J. C. Kinley Co. , 376 S.W.2d 336,
338 (Tex. 1964)). “[I]nformation generally known and readily available is not
protectable, and no trade secret protection is available when the material or
procedure at issue has been publicly disclosed.” Sw. Rsch. Inst. v. Keraplast
Techs., Ltd. , 103 S.W.3d 478, 482 (Tex. App. — San Antonio, 2003, no pet.)).
The word ‘secret’ implies that the information is not generally known or readily available. Courts have refused to give trade secret protection when the material or procedure sought to be protected has been publicly disclosed.
T-N-T , 965 S.W.2d. 18, 22 (Tex. App. Houston [1st Dist.] 1998) (citations
omitted).
CreateAI bore the burden to adequately define its alleged “trade secrets” in a manner sufficient to distinguish the alleged “trade secrets” from
common practice and publicly disclosed information, such as the disclosures
in CreateAI’s extensive patent portfolio . E.g ., Topstone , 729 F. Supp. 3d at 07; see 6.RR.107, 111 16.
Some, like CreateAI’s “sensor - suite,” are physically visible to the public — the arrangement and selection of sensors driven on public roads is
in no way a secret. 5.RR.275 – 76 (Mr. Lu admitting that CreateAI’s “rack”
or “perception housing” can be readily seen from the outside of the truck) .
Others CreateAI chose to make publicly available in its 200+ patents or 400+ patent applications —what CreateAI described as “one of the largest
organic patent portfolios [in AV trucking].” 5.RR.288; 6.RR.111. The breadth
of CreateAI’s patent portfolio around AV technology demonstrates the
conscious choice by CreateAI to forego trade secret protection in favor of
patent protection. 6.RR.108 – 14; 7.RR.224; FMC Techs., Inc. v. Murphy , 679
S.W.3d 788, 807, 813 (Tex. App. — Houston [1st Dist.] 2023, pet. denied)
(information is readily ascertainable and not a trade secret if it is published,
including in patents or patent applications).
CreateAI admitted at the temporary injunction hearing what is already evident from its public patent portfolio — that many of its patents relate to AV
sensor-suites, safety, and decision-making i.e. , its three alleged “trade
secrets.” 5.RR.200 01; see also 6.RR.111 14 (Mr. Lu admitting that
*63 CreateAI’s patent portfolio covers all three categories of the alleged “trade
secrets”).
Further, as Mr. Lu acknowledged and just another example, CreateAI publicly disclosed the locations of its cameras, LiDAR, and radar devices and
the use of a dual CAN-bus architecture in its 2022 Investor Day Presentation.
6.RR.86; 8.RR.124 . And many of the sensors in CreateAI’s sensor -suite were
bought off-the-shelf, meaning they are freely available for use by other
industry participants. See 6.RR.98 99.
Mr. Lu also testified that, despite the admitted overlap, CreateAI has made no effort whatsoever to determine what, if any, of its AV technology is
a trade secret not covered by one of its hundreds of patents. 6.RR.115 16.
That admission alone is sufficient to justify the trial court’s denial of
CreateAI’s requested injunction because it proves that CreateAI’s
inadequately defined “trade secrets” have been publicly disclosed and
cannot, as a matter of law, be protected as trade secrets under TUTSA.
CreateAI’s alleged “trade secrets” are in no way secret. Most (if not all,
it is impossible to know without an adequate definition) have been disclosed,
by CreateAI , to the public at large and those disclosures eliminate its TUTSA
claim —more than justifying the trial court’s order on that basis alone. Even
if there were a sliver of CreateAI’s “trade secrets” that were maintained as
actual secrets under TUTSA, none have been identified by CreateAI. Nor did
CreateAI offer any evidence that Bot Auto has any access to them.
B. CreateAI failed to demonstrate a probable right to recover on its TUTSA claim.
In addition to failing to state a TUTSA claim by failing to adequately define its alleged “trade secrets”— its attempt is too vague and general and its
alleged “trade secrets” are not secret a t all CreateAI also failed to show it
probably would recover on that claim had it properly stated it.
1. CreateAI failed to prove any alleged trade secret was acquired through improper means.
First, it is impossible to acquire public information through improper means. Second, even assuming they were not public, CreateAI nonetheless
offered no evidence that Bot Auto acquired any of its “trade secrets” through
improper means. To the contrary, it conducted two “investigations” that in
an ordinary trade secret case would be expected to result in concrete
evidence but in this case resulted in … nothing.
Indeed, not only was the individual at issue in the laptop investigation, Mr. Wang, neither terminated nor disciplined, he was given accelerated
shares as a retention bonus following the investigation. 6.RR.125 27. And
Mr. Wang’s computer was imaged in full— CreateAI has access to every
download, upload, and use of the computer they claim was put up as a
“smokescreen , ” and yet, CreateAI did not submit any of that into evidence or
rely on any aspect of that forensic analysis to evidence improper acquisition
of its trade secrets by Bot Auto (because there is none). 5.RR.242;
7.RR.83 – 85; CreateAI’s Br. at 47. Instead, CreateAI points to the innocuous
fact that Mr. Wang volunteered his personal computer for review (following
a change in company policy that no longer allowed company work on
personal computers) and asserts, with no support , that the “incident” was
“orchestrated” by Dr. Hou. 5.RR.241 – 42; CreateAI’s Br. at 9 – 11, 24 – 28
(citing to its own application, CR.28, and its own “information and belief”
that “Bot Auto has conspired with Hou, Wang, and other persons to effect
the misappropriation and misuse of [CreateAI ’s ] trade secret information”) .
Similarly, the solicitation “investigation” resulted in no evidence of any wrongdoing by Bot Auto. Instead, the testimony by Dr. Hou evidenced the
opposite: N one of Bot Auto’s employees were recruited directly from
CreateAI and all were required to sign questionnaires and certifications
before joining “to ensure that no confidential or proprietary information is
used at Bot Auto.” 7.RR.68– 76; 5.RR.88 – 89. The evidence shows nothing
improper about Dr. Hou’s (or any other former employee’s) departure. Bot
Auto has put in place policies and procedures to ensure that no trade secrets
migrated from CreateAI to Bot Auto, intentionally or unintentionally.
7.RR.68 76; 5.RR.88 89, 91, 175, 179, 181.
Finally, if CreateAI had any real suspicion that Bot Auto possessed its trade secrets, it could have reviewed Bot Auto’s source code that was made
available prior to the hearing (and remains available today). 6.RR.158 59.
Instead, CreateAI relies on speculation and allegations —which it “supports”
with the same speculation and allegations in its own pleading; of course,
none of this is evidence. Shamoun , 398 S.W.3d at 282; CreateAI ’s Br. at 9–
11, 25 (citing to its own application, CR.24 28); see supra n.8.
After failing to adequately identify and present evidence of actionable trade secrets, it failed to show an improper acquisition of “trade secrets” ; the
trial court’s denial of the injunction can be affirmed on this basis too.
2. CreateAI failed to prove any use or threatened use of an alleged trade secret.
Even if CreateAI had provided evidence of trade secrets and their improper acquisition (which it did not and cannot), denial still was proper
because CreateAI offered no evidence of the third prong use or threatened
use of the alleged trade secrets. CAE Integrated, L.L.C. v. Moov Techs., Inc. ,
44 F.4th 257, 263 (5th Cir. 2022) (affirming denial of an injunction because
an inference of misappropriation based on defendant’s success “is
insufficient to support a finding that [the defendant] used [the applicant’s]
trade secrets.”).
Again, quite the opposite, CreateAI failed to rebut Dr. Hou ’s repeated testimony (a) that Bot Auto has no access to CreateAI ’s trade secrets, (b) that
they would nonetheless be useless to Bot Auto, and (c) why ( i.e. , the
fundamental differences between the two systems he designed — the more
recent of which takes advantage of the latest advancements in AI). 5.RR.61, 69, 83 – 85, 90, 109 – 11, 187 – 88, 190 – 91.
a. CreateAI ’s support — two pre-TUTSA cases — examined evidence of actual trade secret possession by a competitor not present here.
CreateAI relies principally on two cases. CreateAI’s Br. at 34 38 (citing
IAC, Ltd. v. Bell Helicopter Textron, Inc. , 160 S.W.3d 191 (Tex. App. — Fort
Worth 2005, no pet.); T-N-T Motorsports v. Hennessey Motorsports , 965
S.W.2d 18 (Tex. App. — Houston [1st Dist.] 1998)). Both of these cases pre-
date TUTSA’s 2013 enactment, reviewed the issuance (not denial) of a
temporary injunction, and, most important, examined direct evidence of
(1) actual possession (2) of actual trade secrets (3) by a competitor
positioned to use them none of which are present here. See Bell Helicopter ,
160 S.W.3d. at 200 (“When a defendant possesses trade secrets and is in a
position to use them , harm to the trade secret owner may be presumed .”)
(emphasis added).
CreateAI makes no effort to tie the facts of either case to the evidence presented to the trial court here. See CreateAI’s Br. at 34– 39. Instead,
CreateAI cites these pre-TUTSA cases to argue that, because it has trade
secrets and Bot Auto is a competitor (both of which were rebutted at the
hearing) that the other elements of its TUTSA claim (improper acquisition,
threatened use, and injury) should simply be presumed . Id. Of course, that
entire-case presumption is not warranted by the facts or the law, and also is
not supported in any way by Bell Helicopter or T-N-T , which still required
the plaintiff to produce evidence on each element of its trade secret claim. Bell Helicopter, 160 S.W.3d at 200; see also T-N-T , 965 S.W.2d at 24.
In CreateAI ’s first case, Bell Helicopter, a defendant (a repair contractor) gained access to Bell’s proprietary information through a
confidentiality agreement as it contracted to repair rotor blades for Bell’s
206B helicopter. Bell Helicopter, 160 S.W.3d at 196 97. The defendant in
Bell Helicopter then worked with a Bell-competitor (a codefendant) to design
an identical replacement rotor blade for the 206B helicopter, offering it at a
lower price than Bell’s 206B. Id. at 196 200.
In contrast to here, Bell presented actual evidence the defendant possessed and was using the trade secrets at issue the defendant had the
proprietary blade data in its library and the Bell-competitor had access to
it and was actively using it. Id. at 199 200. No such evidence of trade secrets
or use is present here. The presumption of harm in Bell Helicopter was
permitted because Bell showed evidence that a competitor actually possessed
and was using the trade secrets at issue no such evidence was presented
here.
The second case also looks nothing like the present and does not permit the entire-case presumption CreateAI seeks. In T-N-T , Roy Terpstra ran a
part-time body shop (“T -N- T”), designing and manufacturing products for
Hennessey Motorsports (“Hennessey”) as the “director of Viper operations”
that Hennessey used in the high-performance upgrade packages it sold for
Dodge Vipers (called “Venom upgrades”). T-N-T , 965 S.W.2d at 20, 24.
In 1997, Roy Terpstra quit his job at Hennessy and ten days later he “ attended a Viper Club meeting in Florida at which he distributed T-N-T business cards and [] brochures advertising the availability of high grade performance upgrades for Vipers[, claiming] that the T-N-T upgrades were identical to the Venom upgrades, that he had learned how to create these packages as an employee of [Hennessey], and that he offered the same upgrades as [Hennessey] but at a better price[, that] at least three of [Hennesse y’s] customers had already moved their business to T- N-T [and Hennessey] was his only competitor. ”
Id. at 20 21.
*70 The present facts could not be more different Dr. Hou gave extensive testimony on the fundamental differences between the two systems at issue;
how Bot Auto’s was built from the ground -up with no reliance on CreateAI ’s
technology ; and how it does not have and is not using CreateAI’s
information, that would be useless to it anyway. 5.RR.190 91. The court
in T-N-T , on the other hand, found that “[t]he record indicates that
appellants possess appellee’s confidential information and are in a position
to use it to compete directly with appellee.” T-N-T , 965 S.W.2d at 24
(emphasis added).
So while some Texas cases presume harm when there are actual, properly identified and proven trade secrets possessed by a defendant that is
in a position to use them, CreateAI failed to get that far. As explained above,
CreateAI failed to specifically identify trade secrets, failed to identify them in
a way that distinguishes them from common practice and publicly disclosed
information, offered “trade secrets” that are too general , demonstratively
public, and not secret (such as the disclosures in CreateAI’s extensive patent
portfolio), and failed to show how it was likely to recover on the TUTSA claim
it failed to properly state.
*71 While the technologies of both companies involve autonomous trucks, the similarities end there (as Dr. Hou explained) — that alone prevents any
inference of use or harm , as does CreateAI’s failure to identify any
information worthy of trade secret protection or to rebut Bot Auto’s evidence
that it is not competing with CreateAI and is not in a position to use any of
the alleged “trade secrets.” [66] See supra Arg. I.A; Arg. I.B; see also St. Jude
Med. S.C., Inc. v. Janssen-Counotte , No. A-14-CA-00877-SS, 2015 WL
11438611, at *3 (W.D. Tex. Oct. 30, 2015) (“to establish threatened
disclosure, the law requires [plaintiff] to show how disclosure of specific
trade secrets would benefit [defendant]”) (citing Conley v. DSC Commc’ns
Corp. , No. 05 – – 01051 – CV, 1999 WL 89955, at *5 (Tex. App. — Dallas Feb.
24, 1999, no pet.); Cardinal Health Staffing Network, Inc. v. Bowen , 106
S.W.3d 230, 243 (Tex. App. Houston [1st Dist.] 2003, no pet.)).
Turning to the three specific categories of “trade secrets” CreateAI alleges here, Dr. Hou explained that, under CreateAI ’s broad definition, they
all exist in every AV because every AV company has the “same goal” of
*72 (1) allowing sensors to see; (2) implementing effective decision-making that
complies with traffic rules; and (3) safely avoiding injury to people or objects.
5.RR.109 – 11. [67] Despite the ubiquitous nature of CreateAI ’s “trade secrets,”
there is no evidence that Bot Auto is using any of CreateAI ’s proprietary
information; the two systems work toward a goal of autonomous driving, but
in fundamentally different ways below the surface —Bot Auto’s using new AI
technology not even available when CreateAI’s was developed . 5.RR.109 – 11.
b. Bot Auto is not using, and is not in a position to use, CreateAI ’s sensor -suite technology. First, as to CreateAI ’s “sensor -suite, ” the only similarity with Bot Auto’s is that they both use sensors, which is like saying they both have
headlights. [68] Otherwise, disparities abound — rebutting any threat of
potential use by Bot Auto of CreateAI ’s visible, public “sensor - suite”
technology. 5.RR.96, 99 – 108, 123.
From the start, the selection of sensors is different at Bot Auto in fact, there is “no similarity . ” 5.RR. 96, 99 108, 106 (Bot Auto takes advantage of
high-definition LiDARs that were not available when CreateAI ’s sensor -suite
was designed). The arrangement of sensors, as Mr. Lu testified, [69] is different
*73 at Bot Auto. See id. ; 5.RR.107 (Bot Auto is using different makes and models
of sensors and arranging them differently — placing high-resolution LiDARs
very high, “basically at the top of the roof,” while CreateAI places low-
resolution LiDARs very low, “about 1.5 meters” off the ground), 5.RR.100–
01 and 6.RR.123 (showing Bot Auto uses Freightliner trucks manufactured
by Daimler while CreateAI used Peterbilt and International truck brands).
And the processing of sensor information is different at Bot Auto — its neural
network no longer requires sensor fusion to translate data from different
types of sensors. 5.RR.75 76, 100 (Dr. Hou explaining, because the sensor
selection and arrangements differ, so does the data collection, training work,
and training process).
c. Bot Auto is not using, and is not in a position to use, CreateAI ’s decision -making technology.
The fundamental difference between Bot Auto’s technology (that harnesses the latest developments in AI — i.e. , transformers) and CreateAI ’s
antiquated technology (developed before transformers were available) is
seen in the “decision - making” technology that CreateAI claims is a trade
secret. The parties’ fundamentally incompatible systems belie any
speculation of use by Bot Auto, as did the direct, unrebutted evidence. All AI
models make decisions (again, this “trade secret” is a uniform characteristic
of any AV). But how Bot Auto’s pretrained TNN makes decisions is
completely different from CreateAI ’s CNN . See 5.RR.190 – 91; 6.RR.52 – 53, 68.
CreateAI offered no evidence of “use” other than its own speculation that Bot Auto’s timeline for commercialization could only be explained by
using its alleged trade secrets. CreateAI’s Br. at 24 (citing to its own
application). But a defendant’s success does not support an inference of trade
secret misappropriation. CAE Integrated , 44 F.4th at 263 (5th Cir.
2022).
And Dr. Hou testified extensively to the way Bot Auto’s technology was
built from the ground-up, without any reliance on CreateAI ’s antiquated
platform, and how that blank slate enabled it to achieve a shorter, and less
expensive, path to commercialization. 5.RR.78 79 (Dr. Hou explaining the
benefit of starting Bot Auto’s technology from scratch because he could fully
embrace the newest technology without having to incur the expenses and
delays of pivoting he did not have to pay any “managerial debts” or remove
any “lattices”).
Finally, to the extent any testimony on the technology conflicts, accreditation of Dr. Hou’s, as the creator of both systems, over Mr. Lu ’s (who
has no personal experience with or knowledge of Bot Auto’s system and who
joined CreateAI years after its system was developed) is well-supported. [70]
6.RR.82, 120 (Mr. Lu has never typed a single line of CreateAI ’s source code
and has no username for its source code repository); see INEOS Grp. Ltd. v.
Chevron Phillips Chem. Co., LP , 312 S.W.3d 843, 848 (Tex. App. — Houston
[1st Dist.] 2009, no pet.) (citing Davis v. Huey , 571 S.W.2d 859, 862 (Tex.
1978)) (abuse-of-discretion review requires deference to the trial court’s
resolution of conflicting evidence). [71]
d. Bot Auto is not using, and is not in a position to use, CreateAI ’s safety technology.
Similar to the sensor and decision- making technologies, the “safety” technology CreateAI claims is a trade secret really is a “central dogma or
central pursuit” of every AV company. 5.RR.111. The safety features — i.e. ,
redundancies, are a “general concept” that “every component -level failure
should not lead to the failure of the whole system.” 5.RR.111– 12 (like two
engines in an airplane). As with the other “trade secrets,” Bot Auto’s pursuit
of this goal is fundamentally different in application from CreateAI ’s because
*76 it uses a different brand of truck and different steering component, for
example, so the interfacing protocols differ from each other, as do the failure
modes. 5.RR.111 16 (Dr. Hou explaining the safety systems of the two
companies have as much in common as that of a Camry and an Accord both
have fail-safes and redundancies, but they use a completely different
vocabulary in determining the types of failures and each arrive at different
solutions).
Bot Auto demonstrated that it is neither using nor in a position to use any of CreateAI ’s alleged “trade secrets , ” and CreateAI offered no evidence
to the contrary.
C. CreateAI failed to show a probable, imminent, irreparable injury.
CreateAI also failed to show that it faces a probable, imminent, irreparable injury.
While some cases have held that a probable TUTSA claim demonstrates potential harm as a matter of law, they are inapplicable here
because CreateAI failed to produce evidence on any element of its trade
secret claim. See supra Arg. I.A, Arg. I.B.; see Bell Helicopter, 160 S.W.3d at
200 (requiring actual trade secrets and a position to use them before
*77 presuming harm). CreateAI nonetheless relies heavily on these cases ( Bell
Helicopter, T-N-T ) — demanding a presumption of irreparable harm
throughout its brief (and implicitly conceding, and again failing to
demonstrate, it faces any actual harm from the denied injunction).
CreateAI’s Br. at 35 36, 49 –50 (“All of this evidence is amply sufficient,
under [ Bell Helicopter and T-N-T ], to establish the presumption that
[CreateAI] has suffered probable, imminent, and irreparable harm from the
demonstrated misappropriation”) ; see also CreateAI’s Issues Presented
(asking this Court to reverse, not because it met its burden to demonstrate
injury, but because it was entitled to a “presumption of [] harm as a matter
of law ”).
A party seeking injunctive relief must demonstrate irreparable injury — which cannot be established with mere fear or speculation. N. Cypress
Med. Ctr. Operating Co., Ltd. v. St. Laurent , 296 S.W.3d 171, 175 (Tex.
App. — Houston [14th Dist.] 2009, no pet.).
*78 Failure to demonstrate even one aspect of an injury worthy of injunctive relief (probable, imminent, irreparable) is fatal, and CreateAI
failed to satisfy all three.
1. Injury is not probable; CreateAI ’s TUTSA claim fails foundationally for lack of trade secrets.
First, for all the reasons CreateAI ’s TUTSA claim is not adequately stated or likely to succeed, any threatened injury from the improbable claim
is also highly unlikely and purely speculative. Legacy Home Health Agency,
Inc. v. Apex Primary Care, Inc. , No. 13-13-00087-CV, 2013 WL 5305238, at
*2 (Tex. App. — Corpus Christi-Edinburg 2013, pet. denied) (reversing a trade
secret injunction where the applicant had only a “ mere fear or apprehension
of the possibility of injury”) (citing Frey v. DeCordova Bend Estates Owners
Ass ’ n , 647 S.W.2d 246, 248 (Tex. 1983)).
2. Injury is not imminent (TuSimple is now CreateAI).
Second, in addition to lacking any foundation, CreateAI ’s speculative fear of injury lacks any imminence. CreateAI and Bot Auto are not
competitors a key factor for determining the irreparable nature of any
potential harm. Bianco v. Globus Med., Inc. , No. 2:12-CV-00147-WCB, 2014
WL 1049067, at *5 6 (E.D. Tex. Mar. 17, 2014) (surveying trade secret cases
and noting that “the plaintiff’s status as a competitor or noncompetitor of the
defendant weighs heavily in the court’s determination whether to grant
injunctive relief”). CreateAI has pivoted to anime and videogames, even
changing its name and all its branding, including its website. See 5.RR.244;
6.RR.116, 122, 131. This critical fact was immediately clear to the trial court,
who touched on the lack of imminent harm at the initial hearing failing to
understand “what would be different today versus 14 days from now.”
2.RR.37 –39 (struggling “to appreciate the true urgency [as] [i]t’s j ust unclear
to me why this has to happen now.”).
Further, CreateAI ’s own conduct (its industry pivot and delay in bringing and prosecuting this case) shows that it faces no imminent harm. It
pivoted from AV manufacturing (as “TuSimple”) to anime and videogames
(as “CreateAI”) and faces no harm (much less imminent harm) from Bot Auto
succeeding with its new, home-grown technology in CreateAI ’s former
industry while this case proceeds to trial. 6.RR.76 77.
CreateAI also waited several months to bring this lawsuit in October 2024 even though it was aware for more than a year prior that its ousted CEO
was founding another AV trucking company. 5.RR.253; 8.RR.179; Shenzhen
Chengront Tech. Co., Ltd. v. Besign Direct , No. 1:22-CV-10281-JLR, 2022
WL 17741496, at *1 (S.D.N.Y. Dec. 9, 2022) (“several months’ delay in
seeking injunctive relief indicates Plaintiff will not be irreparably harmed by
further delay in getting injunctive relief.”); see also Galtex Prop. Invs., Inc.
v. City of Galveston , 113 S.W.3d 922, 926 (Tex. App. — Houston [14th Dist.]
2003, no pet.).
And after delaying in bringing the case, CreateAI has also delayed its prosecution of it, even seeking a stay — acting with no urgency to remedy the
“imminent” harm it allegedly faces . CreateAI ’s own industry -pivot and
litigation delays show that it faces no imminent harm.
3. Injury is not irreparable (CreateAI is trying to license — monetize — its alleged trade secrets). Finally, any injury CreateAI claims to face is compensable with money damages (as its own efforts to license, and monetize, the purportedly
misappropriated technology demonstrates). 2.RR.42; 6.RR.77 (“We’re
attempting to license.”). A party suffers irreparable injury only if it cannot be
“adequately compensated in damages or the damages cannot be measured
by any certain pecuniary standard.” Mabrey v. SandStream, Inc. , 124
S.W.3d 302, 317 (Tex. App. Fort Worth 2003, no pet.). And a party willing
to license its technology is self-evidently willing to accept money in place of
exclusivity. See Nichia Corp. v. Everlight Elecs. Co. , No. 02:13-CV-702-JRG,
2016 WL 310142, at *66 (E.D. Tex. Jan. 25, 2016), aff’d sub nom. , Nichia
*81 Corp. v. Everlight Americas, Inc. , 855 F.3d 1328 (Fed. Cir. 2017)
(“[Plaintiff’s] licensing of the patents-in-suit to the suppliers of 41% of the
global LED market also precludes a finding of irreparable harm.”). Any
impact on CreateAI’s so-far-failed efforts to monetize its outdated
technology is repairable by money damages. See id.
D. CreateAI failed to carry its trial and appellate burdens. Before the trial court, CreateAI bore the burden to state a TUTSA claim, prove it was probable, and that it faced a probable threat of imminent,
irreparable harm. Hollins , 620 S.W.3d at 405; Butnaru, 84 S.W.3d at 204.
CreateAI failed on all counts and the injunction was properly denied.
On appeal, CreateAI bears an even higher burden of demonstrating that the trial court’s decision (which must be given every reasonable
inference in favor of sustaining it without specific findings and conclusions)
was “so arbitrary that it exceeded the bounds of reasonable discretion.”
Midway CC Venture I, LP v. O&V Venture, LLC , 527 S.W.3d 531, 533 (Tex.
App. Houston [1st Dist.] 2017, no pet.); Younker , 154 S.W.3d at 696.
CreateAI has not and cannot meet this high burden for the extraordinary
relief it seeks —the unrebutted evidence supports the trial court’s denial of
the injunction, and the order should be affirmed.
Finally, the equities likewise favor determining that CreateAI (who failed to sufficiently identify any information worthy of trade secret
protection and shut down its AV operations more than two years ago) was
not entitled to an injunction against Bot Auto (who has already suffered
immense harm from CreateAI ’s bad faith “trade secret” claim). 5.RR.119 20
(Dr. Hou explaining the lawsuit has been raised in every investment meeting
and, at this critical stage of fundraising, even an improper injunction would
kill the company); 8.RR.179.
P RAYER Appellee Bot Auto prays the trial court’s order denying CreateAI ’s application for a temporary injunction be affirmed in all respects and that it
be awarded such other and further relief to which it may be entitled.
Respectfully submitted, By: /s/ Joseph A. Fischer, III Joseph A. Fischer, III, lead counsel tfischer@jw.com Texas Bar No. 00789292 Leisa Talbert Peschel lpeschel@jw.com Texas Bar No. 24060414 Victoria C. Emery temery@jw.com Texas Bar No. 24126228 J ACKSON W ALKER LLP 1401 McKinney, Suite 1900 Houston, TX 77010 Telephone: (713) 752-4530 Facsimile: (713) 308-4130 and Allison Allman aallman@jw.com Texas Bar No. 24094023 777 Main Street, Suite 2100 J ACKSON W ALKER LLP Fort Worth, TX 76102 Telephone: (817) 334-7200 Facsimile: (817) 334-7290 C OUNSEL FOR A PPELLEE B OT A UTO TX I NC .
C ERTIFICATE OF C OMPLIANCE Relying on the word count function of the computer software used to prepare this document, the undersigned certifies that the Brief of Appellee
contains 14,867 words (excluding the sections excepted under T EX . R. A PP . P.
9.4(i)(1)) and was typed in 14-point font with footnotes in 12-point font.
/s/ Allison Allman Allison Allman C ERTIFICATE OF S ERVICE
I hereby certify that on April 23, 2025, a true and correct copy of the foregoing document was served by electronic service on all counsel of record
in accordance with the Texas Rules of Appellate Procedure.
/s/ Allison Allman Allison Allman *85 Automated Certificate of eService This automated certificate of service was created by the efiling system.
The filer served this document via email generated by the efiling system
on the date and to the persons listed below. The rules governing
certificates of service have not changed. Filers must still provide a
certificate of service that complies with all applicable rules.
Yvonne Ferrari on behalf of Joseph Fischer
Bar No. 789292
yferrari@jw.com
Envelope ID: 100000850
Filing Code Description: Brief Not Requesting Oral Argument
Filing Description: Brief Not Requesting Oral Argument
Status as of 4/23/2025 3:16 PM CST
Associated Case Party: Bot Auto TX Inc.
Name BarNumber Email TimestampSubmitted Status
Nicole Burkholder nburkholder@jw.com 4/23/2025 2:53:44 PM SENT
Leisa Peschel 24060414 lpeschel@jw.com 4/23/2025 2:53:44 PM SENT
Leisa Peschel 24060414 lpeschel@jw.com 4/23/2025 2:53:44 PM SENT
Alli Allman aallman@jw.com 4/23/2025 2:53:44 PM SENT
Joseph AFischer, III tfischer@jw.com 4/23/2025 2:53:44 PM SENT
Victoria Emery 24126228 temery@jw.com 4/23/2025 2:53:44 PM SENT
Victoria Emery 24126228 temery@jw.com 4/23/2025 2:53:44 PM SENT
Case Contacts
Name BarNumber Email TimestampSubmitted Status
Joseph Fischer 789292 tfischer@jw.com 4/23/2025 2:53:44 PM SENT
Juana Sauceda jsauceda@jw.com 4/23/2025 2:53:44 PM SENT
Richard Liu richard.liu@consultils.com 4/23/2025 2:53:44 PM SENT
Yvonne Ferrari yferrari@jw.com 4/23/2025 2:53:44 PM SENT
Associated Case Party: TuSimple Holdings, Inc.
Name BarNumber Email TimestampSubmitted Status
Cliff PRiley criley@dykema.com 4/23/2025 2:53:44 PM SENT
Cliff PRiley criley@dykema.com 4/23/2025 2:53:44 PM SENT
Christopher Kratovil ckratovil@dykema.com 4/23/2025 2:53:44 PM SENT
Christopher Kratovil ckratovil@dykema.com 4/23/2025 2:53:44 PM SENT
*86 Automated Certificate of eService This automated certificate of service was created by the efiling system.
The filer served this document via email generated by the efiling system
on the date and to the persons listed below. The rules governing
certificates of service have not changed. Filers must still provide a
certificate of service that complies with all applicable rules.
Yvonne Ferrari on behalf of Joseph Fischer
Bar No. 789292
yferrari@jw.com
Envelope ID: 100000850
Filing Code Description: Brief Not Requesting Oral Argument
Filing Description: Brief Not Requesting Oral Argument
Status as of 4/23/2025 3:16 PM CST
Associated Case Party: TuSimple Holdings, Inc.
Christopher Kratovil ckratovil@dykema.com 4/23/2025 2:53:44 PM SENT
Isaac Villarreal ivillarreal@dykema.com 4/23/2025 2:53:44 PM SENT
Kathy Lowery klowery@dykema.com 4/23/2025 2:53:44 PM SENT
Jennifer Schmitt jschmitt@dykema.com 4/23/2025 2:53:44 PM SENT
Amanda Gordon agordon@dykema.com 4/23/2025 2:53:44 PM SENT
Lina Null lnull@dykema.com 4/23/2025 2:53:44 PM SENT
Lina Null lnull@dykema.com 4/23/2025 2:53:44 PM SENT
Aaron Kotulek akotulek@dykema.com 4/23/2025 2:53:44 PM SENT
Amanda Gordon agordon@dykema.com 4/23/2025 2:53:44 PM SENT
Salvador Robles SRobles@dykema.com 4/23/2025 2:53:44 PM SENT
Jennifer Schmitt jschmitt@dykema.com 4/23/2025 2:53:44 PM SENT
Timothy J.McCarthy TMcCarthy@dykema.com 4/23/2025 2:53:44 PM SENT
Kathy Lowery klowery@dykema.com 4/23/2025 2:53:44 PM SENT
Isaac Villarreal ivillarreal@dykema.com 4/23/2025 2:53:44 PM SENT
[1] Citations to CR.__ or SCR.__ reference the clerk’s record or supplemental clerk’s record, respectively, with the blank identifying the page. Citations to _.RR.__ reference the reporter’s record, with the first blank identifying the volume and second blank identifying the page.
[2] SCR.21 27; Mar. 3, 2025 Notice. xii
[3] Most temporary junctions are resolved within fourteen days of filing and without discovery, based solely on information the plaintiff should have had before filing the petition and seeking the relief. Tex. R. Civ. P. 680 (“Every temporary restraining order … shall expire by its terms within … fourteen days ….”). xiii
[4] CreateAI ’s Issues Presented misstates both the proper standard of review (abuse of discretion, not mere error) and the concrete evidence required in a trade secret case to merit the presumption of irreparable harm it relies on so heavily. Butnaru v. Ford Motor Co ., 84 S.W.3d 198, 204, 211 (Tex. 2002) ( “[w]hether to grant or deny a temporary injunction is within the trial court ’s sound discretion”; no abuse of discretion exists “if some evidence reasonably supports the trial court’s decision”); IAC, Ltd. v. Bell Helicopter Textron, Inc. , 160 S.W.3d 191, 200 (Tex. App. Fort Worth 2005, no pet.) (“When a defendant possesses trade secrets and is in a position to use them , harm to the trade secret owner may be presumed.”) (emphasis added); infra Arg. I.A.1. (failure to identify an actionable “trade secret”) and Arg. I.B.2 (failure to show a position to use an actionable “trade secret” ).
[5] See infra Arg. I.A.1.
[6] See infra Arg. I.B.2.
[7] See infra Arg. I.C. xv
[8] C reateAI’s Br. at 24 (citing only to allegations in its application, CR.27). CreateAI’s repeated citations to its own application which is not evidence — emphasize its lack of support. Shamoun & Norman, LLP v. Yarto Intern. Group, LP , 398 S.W.3d 272, 282 (Tex. App. —Corpus Christi 2012, pet. dism’d) (pleadings, even if sworn, are not evidence for a temporary injunction). Indeed, every citation CreateAI offered to support its statement of facts refers back to the allegations in its own application for a temporary injunction. CreateAI’s Br. at 16– 26 (citing only CR.6 39).
[9] CreateAI’s Br. at 39.
[10] Id .
[11] See infra n.46.
[12] 5.RR.51 64, 64 (Dr. Hou’s goal at Bot Auto “is to commercialize autonomous driving, using the cutting edge technology that w[as] not available, not even invented when I was in my last career [at TuSimple]. With this new technology, we’re going to make autonomous driving a moneymaking business and commercialize it.”).
[13] Although this Brief generally refers to Appellant as “ CreateAI ,” reflecting its current name and completed pivot from the AV industry, this section will at times refer to it by its former name, “ TuSimple, ” so the evolution and history is clear. SCR.21 27.
[14] As he did with the “bot auto” domain, Dr. Hou created TuSimple before deciding the entity’s purpose. 5.RR.60, 63, 152 – 56 . (“ I do have a habit of registering interesting domain names. … I just, from time to time, collect domain names, yes.”).
[15] Hub-to-hub refers to transportation between two logistic hubs — traveling from a surface street to the highway and back to another logistics hub. 5.RR.62, 158 – 59; CR.665.
[16] Level-4 automation refers to a system that drives itself in limited service areas without the need for a human driver in the vehicle. CR.667.
[17] As both TuSimple and Bot Auto testified at the hearing, the architecture of TuSimple’s neural network (convolutional) was expensive and time-consuming to build and maintain because it had to be manually trained by humans first collecting miles and miles of data, and then physically annotating the data by drawing a rectangle around different objects in order to “teach” the model to recognize a “person” or a “car . ” 6.RR.52 53; 66 68; see infra Arg. I.C.1.
[18] A TuSimple/Hydron partnership was first explored by Mr. Lu as CEO. 8.RR.175; 5.RR.144.
[19] 5.RR.93 94; 6.RR.133, 169.
[20] 5.RR.247 –48 (“morale was shaky, to say the least”); see 8.RR.177 78.
[21] CR.1112; 7.RR.522; 5.RR.57.
[22] TuSimple Holdings, Inc. (now CreateAI Holdings, Inc.) is the parent company of two different entities the U.S. entity, TuSimple, Inc. (which has no active AV operations) and the China entity, TuSimple China (which has pivoted to anime and videogames). SCR.21 27. The now-shuttered U.S. entity is the only company referenced in the Employee Proprietary Information and Inventions Agreement with Dr. Hou. 7.RR.77 82.
[23] 5.RR.244 , 245 (“we’re fighting for survival”), 247 (“there w[ere] very bad articles about us in the press. The leadership was uncertain.”) ; 6.RR.116, 122, 131.
[24] The purchase of the bot.auto domain in December 2022 was unrelated to Dr. Hou’s creation of a new business in the spring of 2023. See 5.RR.60, 152 –56. (“I do have a habit of registering interesting domain names. … I just, from time to time, collect domain names, yes.”).
[25] 5.RR.57 –61, 64 (“artificial intelligence has fundamentally changed,” the “ecosystem has greatly matured,” and Cha t GPT is based on technology that runs on a “fundamentally upgraded” neural network architecture— i.e. , the transformer or language model).
[26] 5.RR.56, 69 (“at that time [when TuSimple’s system was created], the new paradigm of neural network learning ha[d] not been invented.”).
[27] The “T” in ChatGPT stands for “transformer , ” as in “Generative Pre -trained Transformer.” 5.RR.69.
[28] 5.RR.67 69 (Dr. Hou explaining that “in TuSimple we used very cheap laborers in China to basically draw a box around every like draw a rectangle in image to denote this is a person, that is a car.”).
[29] In 2022, a company called NVIDIA released an advanced processor (“ GPU ”) that made the promise of a pre-trained transformer network possible. See 5.RR.73; 8.RR.186.
[30] 5.RR.64 –65, 69 (“at that time, the new paradigm of neural network learning ha[d] not been invented.”) , 161 –62 (“I’m very sure that there is no transformers in the
[32] 5.RR.190 (“Any data that TuSimple has collected are using a different configuration of the sensors. That data format is not even the same, so we cannot even use them. So it has zero value to Bot Auto.”); 5.RR.90 (explaining TuSimple’s source code would not work in Bot Auto’s system as it would be like “having a kidney from a pig [] implanted into me, it just won’t work.”).
[33] As Mr. Lu acknowledged at the temporary injunction hearing, AV technology “is constantly iterating” so two years is a lifetime in t he industry, and you are quickly left behind if you stop developing technology with the newest tools (as CreateAI has). 5.RR.259; 6.RR.10, 15 (noting the “constant innovation in sensors” and “constant innovation in terms of the power to compute”).
[34] 8.RR.179 80.
[35] CreateAI’s CEO Mr. Lu recognized the visible difference between the two companies ’ sensor-suites at the temporary injunction hearing. 5.RR.282 (“It’s different, because … there’s like two parts in the middle that’s cut out. … this actually isn’t as well built.”).
[36] CR.647 – 55; 5.RR.111 – 13 (Dr. Hou explaining safety is a “central dogma of central pursuit” for every AV company, to ensure that a “component -level failure should not lead to the failure of the whole system”).
[37] CR.647 55; 5.RR.112 (Dr. Hou explaining redundancy is a “general concept”).
[38] Daimler, for example, advertises redundancy in its braking, steering, power supply, communication, and decision-making systems. CR.657 63.
[39] Dr. Hou cited “the drama between Tesla and Xpeng” as one reason Bot Auto is “very careful we do not accept any injection of exterior data or source code into our repository.” 5.RR.87.
[40] Even Mr. Lu agrees that some of the former TuSimple employees who now work for Bot Auto likely were laid off by TuSimple. 6.RR.122.
[41] 2.RR.37 (citing Camp v. Shannon , 348 S.W.2d 517, 519 (Tex. 1961) (“Writs of injunction should not issue on mere surmise.”) ).
[42] CR.173 – 76, 863.
[43] CR.6 – 172, 517 – 835; 1003 – 21; 1055 – 64; 1074 – 90.
[44] 5.RR.51 192 (Dr. Hou’s testimony); 5.RR.192 6.RR.149 (Mr. Lu’s testimony).
[45] 5.RR.61 ( “ Q: … Is Bot Auto using in any way any trade secrets of TuSimple for the furtherance of Bot Auto? A: No. ” ); 5.RR.83 84 ( “Q: Do you have any aspect of TuSimple’s source code? A: No, I do not. Q: If you had any aspect of TuSimple’s source code, would it be useful to you at Bot Auto? A: No, it would not be useful … .” ); 5.RR.109 11 ( “Q: … Are you using trade secrets of TuSimple as it relates to the sensor suite? A: Not at all.
[48] Likewise, the TuSimple/CreateAI documents that mentioned “Kubernetes” were made “in the middle of 2023,” months after Dr. Hou left. 6.RR.56 57, 93 – 96.
[49] Mr. Lu also mentioned for the first time an “MVP” project—the “next generation” of sensor-suite design. 6.RR.34 35. But that “iteration of designs” “was evolving” (more of a concept of a plan) and Mr. Lu testified that it was never put on a truck in the U.S. 6.RR.106 –07 (“I do not believe we put it on a truck in the U.S., correct.”).
[50] See Hilb, Rogal & Hamilton Co. v. Wurzman , 861 S.W.2d 30, 35 (Tex. App. Dallas 1993, no writ) (criticizing plaintiff for delaying trial with “unnecessary” appeal, “like many temporary injunction appeals”; and “reaffirm[ing] our conclusion that the fastest way of obviating the hardship of an unfavorable preliminary order is to try the case … ”).
[51] CreateAI also filed its brief here on the conventional schedule even though its appeal is accelerated CR.1151 –54; Jan. 13, 2025 Docketing Stmt. at 3; Court’s Jan. 30, 2025 Letter.
[52] E.g. , Bell Helicopter , 160 S.W.3d. at 197; T-N-T , 965 S.W.2d. at 23 24.
[53] See T-N-T , 965 S.W.2d. at 22 (“The word ‘secret’ implies that the information is not generally known or readily available. Courts have refused to give trade secret protection when the material or procedure sought to be protected has been publicly disclosed.”) (citations omitted).
[54] Bell Helicopter , 160 S.W.3d. at 200 (“When a defendant possesses trade secrets and is in a position to use them … .”) (emphasis added).
[55] Id. at 197 (“An injury is irreparable if the injured party cannot be adequately compensated in damages or if the damages cannot be measured by any certain pecuniary standard.”).
[56] See infra n.8 regarding CreateAI’s repeated citations to its application as “support.”
[57] CreateAI’s Br. at 32, 34 35, 38 50.
[58] Bell Helicopter , 160 S.W.3d at 194 (emphasis added).
[59] What little Bot Auto can decipher from CreateAI’s vaguely-defined and highly- abstracted trade secrets seems to raise even more issues fatal to the injunction application, in addition to the secrecy issue discussed below: (1) it doesn’t appear CreateAI even owns the “trade secrets” at issue ( most, if not all, were contractually assigned to a different entity (7.RR.77 – 82) and many belong to truck manufacturing companies, not CreateAI (CR.558 – 59, 669 – 75)); (2) most of the documents with “trade secret” information CreateAI presented at the hearing were created in 2023, months after Dr. Hou was terminated in 2022 (7.RR.95 193, discussed at 5.RR.208 – 09 as “includ[ing] extensive trade secret information” that CreateAI shared with third party investment bankers “to sell parts or all of the U.S. operations”) (7.RR.403– 83 discussed at 6.RR.93 – 96 as having been made “in the middle of 2023” months after Dr. Hou left); and (3) any suspected misappropriation of “Covered IP” under the NSA, as CreateAI alleges here, would have required CreateAI to report the violation to CFIUS within three days, which it has not done (CR.552 53).
[60] See supra I.C.2.
[61] Courts require the same specificity for trade secrets alleged under TUTSA’s federal corollary the Defend Trade Secrets Act. B road references to “designs,” “systems,” and “details,” as CreateAI alleged here, are not enough to state a probable trade secret claim. See, e.g. , SL Montevideo Tech., Inc. v. Eaton Aerospace, LLC , No. Civ. 03-3302 RHK/FLN, 2006 WL 1472860, at *3 (D. Minn. May 26, 2006) (“the design of its motor as a whole” cannot be a trade secret), aff ’ d on other grounds , 491 F.3d 350 (8th Cir. 2007); Imax Corp. v. Cinema Techs., Inc. , 152 F.3d 1161, 1167 (9th Cir. 1998) (“dimensions and tolerances” cannot be trade secrets where the plaintiff “failed to indicate precisely which dimensions and tolerances”); ECT Int’l, Inc. v. Zwerlein , 597 N.W.2d 479, 483 (Wis. Ct. App. 1999) (finding plaintiff’s “allegation that the ‘software file system for promis.e is a trade secret’ does not describe which of the components of ‘promis.e’ were misappropriated”).
[62] 5.RR.14, 22 (describing “know - how” as a category of its alleged trade secrets that “is the detailed business of how stuff you need to make actually gets built.”), 5.RR.201, 214, 217, –23, 258, 276, 277, 282, 288; 6.RR.8, 17, 20 (describing “know - how” as implementing its design and architecture), 5.RR. 21 (agreeing that “know -how is all the things you learn about how to build that device and keep it working and perhaps even improve upon it.”), 5.RR.22 23, 42 43, 46, 78, 145, 148; see also 6 .RR.166, 171 (Bot Auto’s counsel pointing out that the “know - how” discussed at length during Mr. Lu’s testimony “is not a trade secret under TUTSA.”).
[63] See supra images at I.C.2.a. And even if there was some aspect of CreateAI’s sensor- suite that was kept secret, it nonetheless differs from the sensor-suite utilized by Bot Auto, even visually, as Mr. Lu testified. 5.RR.282 (“It’s different, because … there’s like two parts in the middle that’s cut out.”).
[64] Indeed, so lacking was the proof of use, CreateAI did not even bother to review Bot Auto’s source code. 6.RR.158– 59.
[65] See supra Arg. I.A; Arg. I.B.
[66] To infer threatened use by a competitor, the products or processes must be substantially similar, not merely occupy the same niche or perform the same general function. Spear Marketing, Inc. v. BancorpSouth Bank , 791 F.3d 586, 601 02 (5th Cir. 2015); Phazr, Inc. v. Ramakrishna , No. 3:19-CV-01188-X, 2020 WL 5526554, at *5 (N.D. Tex. Sept. 14, 2020) (Starr, J.) (“After all, a Ford Pinto and a McLaren 620R perform the same general function of moving a person from one place to another. More is needed [to state a plausible TUTSA claim], and actual similarity must exist between the two products.”) (citing Spear , 791 F.3d at 601 02).
[67] See also 5.RR.96 (explaining all AVs have sensors, just as they all have wheels).
[68] 5.RR.96 (Dr. Hou testifying there is “no similarity” between the sensor -suites); see also 5.RR.99 108, 123.
[69] 5.RR.282 (“It’s different, because … there’s like two parts in the middle that’s cut out.”).
[70] Mr. Lu’s testimony on the alleged similarities between the systems could have been discounted or disregarded because he lacks personal knowledge and also because, at most, even if credited it shows only that CreateAI’s Asia-Pacific entity was exploring the use of transformers in China and the technology was migrated over to the U.S. entity after Dr. Hou’s 2022 termination. See 6.RR.56, 93 – 96, 132. The technology was never put on a CreateAI truck in the U.S. 6.RR.34 35, 106 07; 5.RR.161 – 62.
[71] See also Amend v. Watson , 333 S.W.3d 625, 627 (Tex. App. — Dallas 2009, no pet.); Shor v. Pelican Oil & Gas Mgmt., LLC , 405 S.W.3d 737, 748 (Tex. App. Houston [1st Dist.] 2013, no pet.).
[72] CreateAI claims that Bot Auto’s steering system requirements document uses the same “key parameters” (CreateAI’s Br. at 42– 43) but that document shows the values Bot Auto was considering differ from CreateAI’s. Compare 8.RR.132 with 8.RR.136.
[73] See, e.g. , Harbor Perfusion, Inc. v. Floyd , 45 S.W.3d 713, 716 (Tex. App. Corpus Christi 2001, no pet.); Wash. DC Party Shuttle, LLC v. IGuide Tours, LLC , 406 S.W. 3d 723, 742 (Tex. App. —Houston [14th Dist.] 2013, pet. denied) (“An injunction is not proper when the claimed injury is merely speculative; fear of injury is not sufficient to support a temporary injunction.”) ( citations omitted).
[74] supra n.51 & 52.