Case Information
UNITED STATES DISTRICT COURT WESTERN DISTRICT OF WASHINGTON AT SEATTLE
KAELI GARNER, et al ., Case No. 2:21-cv-00750-RSL Plaintiffs, ORDER GRANTING IN PART PLAINTIFF’S v. MOTION FOR CLASS AMAZON.COM, INC., et al ., CERTIFICATION
Defendants.
This matter comes before the Court on “Plaintiffs’ Motion for Class Certification.” Dkt.
# 257. Having reviewed the memoranda, declarations, and exhibits submitted by the parties, [1] the Court finds as follows:
I. B ACKGROUND
Defendants developed Alexa, a voice-activated, cloud-based service that can access online resources to respond to inquiries and execute commands. When Alexa’s microphones pick up a “wake” word, the device turns on a blue light and begins to stream audio to Amazon’s cloud. The audio is transcribed, and the resulting text strings are processed by an algorithm to generate instructions to the Alexa system about how to respond to the user. After a request has been made, processed, and responded to, defendants store a copy of the audio file, the transcription, the instructions, and associated metadata in an internal Amazon database. The recordings are used to train defendants’ speech-recognition and natural-language-understanding systems, a process that involves human review of approximately 0.01% of Alexa interactions. At this time, customers have the ability to delete recordings, to choose not to have their recordings stored at all, and/or to opt out of human review.
Sometimes Alexa begins streaming when a wake word was not, in fact, used. These events are called “false wakes.” When the audio stream hits the cloud, a cloud-side wake word verification process attempts to confirm the presence of a wake word. If it cannot, the audio stream is terminated, the recording is labeled “not intended for Alexa,” and the system saves “a truncated snippet of audio, i.e ., only enough to determine what triggered the device to activate, . . . both for use in improving the Alexa service and wake word models and to provide users with visibility into what the Alexa device recorded when the streaming indicator was activated.” Dkt. # 310 at ¶ 27.
Plaintiffs assert that both the permanent storage of Alexa interactions and the false wakes are intentional design elements of the service, used to amass huge numbers of voice recordings that can be fed into algorithms and machine learning platforms for continuous improvement training. By 2020, Alexa was processing 2.8 billion communications a week, and defendants retained and continue to use the recorded data to train Alexa and the next generation assistants. Plaintiffs further assert that Amazon failed to provide sufficient notice of the recording, saving, human review, and/or use of the audio captured by the Alexa service in violation of six states’ wiretap laws or Washington’s Consumer Protection Act (“CPA”). They seek certification of six classes of non-registrants, defined as residents of California, Florida, Maryland, New Hampshire, Pennsylvania, or Washington who never registered an Alexa device and resided with a registrant while an Alexa device was active and operational, to pursue wiretap claims under state law. The non-registrant classes seek to recover statutory damages. Plaintiffs also seek certification of a nationwide class of everyone who registered one or more Alexa devices to pursue a CPA claim for unfair or deceptive trade practices. The registrant class seeks (a) compensation for the value of the false wake recordings defendants have compiled and (b) injunctive relief.
Defendants argue that the named plaintiffs’ claims are not viable because they consented to the recordings at issue and were not harmed by the recording, storage, and/or use of communications not intended for Alexa ( i.e ., false wakes). They also argue that all of the proposed classes are too broad in that they sweep in putative class members who were never recorded at all and therefore lack Article III standing. Defendants challenge Kaeli Garner’s and Ronald Johnson’s ability to represent a class of California or Pennsylvania non-registrants, respectively, because both of them registered an Alexa-enabled device. Finally, defendants maintain that the common questions of law and fact raised by plaintiffs’ claims do not predominate over questions affecting only individual members of the putative class and that class treatment would be unmanageable and is therefore not superior to individual litigation.
II. T HRESHOLD I SSUES
A. Viability of Named Plaintiffs’ Claims
Defendants opted not to support their challenge to the viability of the named plaintiffs’ claims in response to the motion for class certification, instead filing a separate motion for summary judgment that did not become ripe for consideration until months after the certification issue was ready to be resolved. The Court declines the invitation to first rule on the summary judgment motion.
B. Article III Standing
Defendants argue that the putative classes are overbroad because they contain class
members who were never recorded and therefore lack Article III standing to pursue either a
wiretap or a CPA claim. Dkt. # 28 at 22. Article III, § 2 of the United States Constitution limits
the federal judicial power to “cases” and “controversies.” Standing is a judicially-created
doctrine designed to protect the separation of powers concerns embodied in the text of Article
III.
Allen v. Wright
,
3 To satisfy Article III standing, plaintiffs must demonstrate that they have (1) suffered an
4 5 injury in fact, which is (2) fairly traceable to the challenged conduct of the defendant, and 6 (3) likely to be redressed by a favorable judicial decision. Lujan v. Defenders of Wildlife , 504
U.S. 555, 560-61 (1992). An injury in fact exists where the plaintiff suffered “an invasion of a legally protected interest” that is “concrete and particularized” and “actual or imminent, not conjectural or hypothetical.” Id. at 560.
The Ninth Circuit has clarified that in class actions seeking injunctive or other equitable
relief, only one plaintiff is required to demonstrate standing.
Olean Wholesale Grocery Coop. v.
Bumble Bee Foods LLC
,
The proposed classes each contain tens of millions of people. Defendants argue that some of those people use Alexa through a push-to-talk button (rather than through a wake word), may have gifted the device to someone else, or may be non-verbal so that they could not have had their voices recorded. They also argue that 900,000 Alexa users have taken advantage of a “do- not-save recordings” function. Putting aside the unknown efficacy of the “do-not-save recordings” function, the option did not exist until September 2020, and there is no reason to assume that the 900,000 were not recorded before that date. Nor is there any information from which the Court could hazard a guess regarding the number of Alexa users who use only push- to-talk, are disabled, or never used a device they were given or purchased. Plaintiffs estimate, based on defendants’ alleged design preferences in favor of over-recording, the business need for voice recordings, and defendants’ calculated “false wake rate,” that Alexa-enabled devices record communications that were not intended for Alexa at least once a month. Even if there are some people who own or live with an Alexa device who were never recorded, there is no reason to suspect, much less presume, that a significant or more than de minimis portion of the proposed classes escaped recording. The problem of individualized standing issues will be considered under the predominance analysis of Rule 23(b)(3).
III. P REREQUISITES OF A C LASS
The party seeking certification must demonstrate that all four prerequisites of Rule 23(a)
are met, as well as one of the requirements of Rule 23(b).
Zinser v. Accufix Research Inst., Inc.
,
A. Numerosity
Rule 23(a)(1) tests whether “the class is so numerous that joinder of all members is impracticable.” Defendants do not contest numerosity.
B. Commonality
Commonality requires “questions of law or fact common to the class.” Fed. R. Civ. P.
23(a)(2). Although the plain text of the Rule constrains itself to searching for common
“questions,” the key inquiry is actually whether the class proceeding will “generate common
answers
apt to drive the resolution of the litigation.”
Dukes
,
The Court finds that the following questions are shared in common by each non- registrant class and will likely generate a common answer for the entire class:
1. whether plaintiffs’ common evidence establishes that defendants intended to and knew that Alexa intercepted communications of non-registrants that were not meant for Alexa; 1 and
2 2. whether Alexa devices are wiretapping “devices” as defined by the applicable state 3 wiretapping statutes.
4 5 With regards to the registrant class, the following common questions are likely to generate 6 common answers:
7 1. whether defendants adequately disclosed information regarding their interception,
8 9 indefinite storage, and use of registrants’ recordings (including conversations not 10 intended for Alexa);
11
2. whether defendants’ failure to adequately disclose information regarding the interception, indefinite storage, and use of registrants’ recordings was an unfair or deceptive act or practice;
3. whether defendants’ unfair and deceptive acts or practices impacted the public interest; 4. whether registrants’ valuation of the utterances defendants compiled without compensation is reasonable and whether the loss of that value is an injury to each class members’ business or property; and
5. whether registrants have demonstrated that defendants’ actions in using intercepted recordings to train its algorithms and machine learning platforms should be enjoined.
The Court finds commonality satisfied in this case.
C. Typicality
Typicality asks whether “the claims or defenses of the representative parties are typical of
the clams or defenses of the class.” Fed. R. Civ. P. 23(a)(3). “The purpose of the typicality
requirement is to assure that the interest of the named representative aligns with the interests of
the class.”
Wolin v. Jaguar Land Rover N. Am
.,
LLC
,
15 Defendants argue that Kaeli Garner and Ronald Johnson are not part of the classes they
16
17
seek to represent and therefore do not “possess the same interest and [did not] suffer the same
18
injury as the [absent] class members.”
Dukes
,
20 21 suggesting that she had purchased and registered an Alexa-enabled Fire TV stick in late 2017. 22 Dkt. # 302-8 at 8 and 11. Defendants argue that she is therefore a registrant, that she is bound by 23 a Washington choice-of-law provision contained in the terms and conditions to which she
24 25 assented, and that she cannot assert claims under California law on behalf of herself or the non- 26 registrant class. It appears that Ms. Garner is not typical of the California non-registrant class 27
28
because she is, in fact, a registrant and subject to the terms and conditions to which registrants agree. Because Ms. Garner is the only plaintiff offered as a representative for non-registrants located in California, that class cannot be certified.
With regards to Mr. Johnson, he seeks to represent a class of Pennsylvania non- registrants. Mr. Johnson acknowledged at his deposition that he had an Alexa app on his phone at some point in time. Dkt. # 302-10 at 4. Defendants provided evidence that use of the Alexa app requires consent to the same terms and conditions of use as the registrants whose wiretap claims have already been dismissed by the Court. Dkt. # 310 at ¶¶ 60-62; Dkt. # 91 at 16-17. Mr. Johnson is not, therefore typical of other non-registrants in Pennsylvania. In the absence of a representative, a Pennsylvania class of non-registrants cannot be certified.
D. Adequate Representation
The representative parties—meaning the named plaintiff as well as class counsel—must “fairly and adequately protect the interests of the class.” Fed. R. Civ. P. 23(a)(4); see also Fed. R. Civ. P. 23(g)(4). Except with regards to Ms. Garner and Mr. Johnson, discussed in the preceding section, defendants offer no reason to suspect that absent class members will not be adequately represented.
IV. M AINTENANCE OF A C LASS
In addition to meeting all four prerequisites of Rule 23(a), a plaintiff must also meet one of the three requirements of Rule 23(b). The Court considers this motion under Rule 23(b)(2) and (b)(3).
A. Rule 23(b)(2)
Plaintiffs seek certification of a class of registrants to pursue injunctive relief under the CPA. Certification under Rule 23(b)(2) is available when defendant “has acted or refused to act on grounds that apply generally to the class, so that final injunctive relief or corresponding declaratory relief is appropriate respecting the class as a whole.” Plaintiffs argue that defendants acted uniformly in relation to the registrant class by recording them in the absence of a wake word and, without disclosure, transcribing, storing, and using those recordings indefinitely for their own benefit. They request that the Court enjoin defendants from using class recordings or data from those recordings to train algorithms and programs and to require defendants to destroy existing recordings, transcripts, and related data.
Defendants argue that a Rule 23(b)(2) class cannot be certified because the primary relief
sought through the CPA claim is monetary damages. While it is clear that Rule 23(b)(2) “does
not authorize class certification when each class member would be entitled to an individualized
award of monetary damages,”
Dukes
,
The Court finds that a Rule 23(b)(2) class of registrants may be certified for the purpose of seeking injunctive relief.
B. Rule 23(b)(3)
An “adventuresome innovation” created specifically with damages class actions in mind,
Rule 23(b)(3) imposes two additional requirements on a putative class beyond the four
prerequisites of Rule 24(a), namely predominance and superiority.
Amchem Prods., Inc. v.
Windsor
,
3 1. Predominance
4 5 Under the Rule 23(b)(3) predominance inquiry, the Court must find “that the questions of 6 law or fact common to class members predominate over any questions affecting only individual 7 members.” Fed. R. Civ. P. 23(b)(3). “An individual question is one where members of a
8 9 proposed class will need to present evidence that varies from member to member, while a 10 common question is one where the same evidence will suffice for each member to make a prima 11 facie showing or the issue is susceptible to generalized, class-wide proof.” Tyson Foods , 577
12 13 U.S. at 453 (quoting 2 W. Rubenstein, N EWBERG ON C LASS A CTIONS § 4:50, 196-197 (5th ed. 14 2012 (internal quotations marks and alteration omitted))). Although somewhat similar to the 15 commonality prerequisite of Rule 23(a)(2), commonality and the predominance requirement of
16
17
Rule 23(b)(3) must be analyzed separately.
Comcast Corp. v. Behrend
,
U.S. at 624 (finding that even if named and unnamed plaintiffs’ shared asbestos exposure was
sufficient to satisfy commonality, too many individual issues regarding the circumstances and
subsequent effects of exposure predominated to uphold class certification). Where individual
inquiries would overwhelm common issues, certification under Rule 23(b)(3) is inappropriate.
Amchem
,
Defendants argue that the issues that will require individual presentations of evidence and judicial determinations outnumber and outweigh the issues that can be resolved in one stroke for each of the proposed classes. In particular, they argue that the non-registrant plaintiffs will have to produce individual evidence that they are part of the class, that they are not bound by the terms and conditions that bind the registrants with whom they lived, that they did not consent to the interception of their communications, that those communications constitute “private conversation” under the relevant wiretap statutes, and that they had a reasonable expectation that their communications would not be recorded. Defendants further argue that every member of the non-registrant class will have to prove the number of wiretap violations in order to claim statutory damages. With regards to the registrant class, defendants maintain that proof of injury and causation will be individualized and that there is no viable method for ascertaining damages across the entire class. Each of these arguments and, if necessary, the standing issue will be considered, separately and collectively, below.
a. Non-Registrant Class Members
i. Identification of Non-Registrant Class Members With regards to the non-registrant class, defendants argue that there is no way to identify all persons within a given state who never registered an Alexa device but who resided with a registrant while an Alexa device was active and operational. This issue is less about predominance and more about manageability. See Briseno v. ConAgra Foods, Inc. , 844 F.3d 1121, 1126-28 (9th Cir. 2017); Wetzel v. CertainTeed Corp ., No. C16-1160JLR, 2019 WL 3976204, at *19 (W.D. Wash. Mar. 25, 2019). The class definition sets forth objective criteria that allows the parties and the Court to recognize any given class member and to evaluate the viability of the claims asserted on their behalf. Whether the absence of a database containing all non-registrants’ names or the related difficulties in providing class notice makes representative litigation unmanageable is an issue to be considered when determining superiority.
ii. Agency and Estoppel Defendants argue that non-registrants may be bound by Amazon’s terms and conditions through an agency or estoppel theory, a determination that can be made only after individualized discovery and fact-finding. With regards to agency, defendants assert that non-registrant spouses, children, and roommates (the purported principals) may have authorized the registrants with whom they lived to register an Alexa device and enter into the terms and conditions on their behalf.
A principal is generally bound by the acts of his or her agent which, based on the
objective manifestations of the principal, are within the express or apparent authority granted to
the agent.
See generally Steadman v. Green Tree Servicing, LLC
, No. 2:14-cv-00854-JLR, 2015
WL 2085565, at *6 (W.D. Wash. May 5, 2015) (citing
King v. Riveland
,
Defendants’ estoppel argument is based on the fact that non-registrants used the Alexa
device in their home and are therefore equitably estopped from avoiding the terms and
conditions to which the registrant agreed, citing
Tice v. Amazon.com, Inc
., No. 5:19-cv-1311-
SVW-KK,
iii. Consent The wiretap statutes of California, Florida, Maryland, New Hampshire, Pennsylvania, and Washington require that plaintiffs prove that their private conversation was intercepted without their consent. Defendants argue that some purported non-registrants may have expressly consented to Amazon’s terms and conditions of use (including that their conversations would be recorded) by buying an Alexa device on amazon.com and/or by signing into the Alexa app. As discussed above, Mr. Johnson, the proposed representative of a class of non-registrants who are pursuing a wiretap claim under Pennsylvania law, had the Alexa app on his cell phone for some unspecified period of time and therefore agreed to the same terms and conditions of use as the registrants whose wiretap claims have already been dismissed by the Court. Dkt. # 302-10 at 4; Dkt. # 310 at ¶¶ 60-62; Dkt. # 91 at 16-17. Plaintiffs argue that (1) the wiretap classes are defined to include only people who did not register an Alexa device and (2) Amazon already possesses information sufficient to challenge the inclusion of a given non-registrant in the non- registrant class. The record shows, however, that registering an Alexa device is not the only way a consumer consents to the recordings, so merely looking at whether an individual has registered an Alexa device does not answer the consent question. That defendants will not have to undertake extensive discovery to identify putative non-registrants who have nevertheless consented does not change the fact that the inquiry is individualized and must be resolved before either liability or damages can be adjudicated.
Defendants also argue that publicly available disclosures, news reports, lawsuits, and
other sources alerted non-registrants to the fact that if an Alexa device awoke – intentionally or
unintentionally – the subsequent communications would be recorded, giving rise to the non-
registrants’ implied consent to the recordings. This issue might, as plaintiffs suggest, be
susceptible to classwide resolution if, after a review of all of the relevant disclosures and
sources, the fact-finder determines that the available information would not have put a
reasonable consumer on notice that he or she was consenting to the recordings.
See Calhoun v.
Google, LLC
,
iv. Content of Recording The wiretap statutes of Washington and Maryland prohibit the interception of “conversations.” RCW 9.73.030; Md. Code Cts. & Jud. Proc. § 10-401(13). While the Court declines to undertake an exhaustive analysis of what is and what is not a “conversation” for purposes of these statutes at this stage of the litigation, unintelligible sounds, random noises, and words with little or no context likely fall outside the definition of “conversation.” Defendants have provided at least two recordings that contain no words, convey no information, and would not likely be considered “conversations.” Dkt. # 294-5 and # 294-7.
Plaintiffs assert that each member of the non-registrant classes was injured and damaged based on the facts that (a) they lived in proximity to an active Alexa device and (b) Amazon’s internal data shows that false wakes were so frequent as to make it almost certain that each non- registrant was recorded at some point. Each of the proposed class representatives have identified at least one recorded communication that would be considered a private conversation. Dkt. # 257 at 13-14. Defendants’ assertion that, amongst all the recordings it has transcribed, maintained, and used, a few of them are unintelligible does not invalidate any named plaintiff’s claim. At most, individualized proof that certain recordings did not capture conversations might impact the calculation of the per violation damages for the Washington and Maryland classes, but the need for some individualized damage calculations does not outweigh the common issues identified by plaintiffs.
v. Reasonable Expectation of Privacy Most of the state wiretap laws at issue require a showing that plaintiff had a reasonable expectation that the communication was private and/or that it would not be intercepted. The analysis involves an objective inquiry which, plaintiffs argue, will be decided for the entire class based on common evidence. Plaintiffs point out that non-registrants were recorded in their homes, a place where one normally expects privacy, and they rely on survey evidence showing that consumers do not like the idea that their in-home communications are being intercepted by a constantly listening device. Defendants, for their part, offer a different survey suggesting that the knowledge and expectations of Alexa users vary widely, with more than half understanding that Alexa was recording and keeping communications. To the extent a given non-registrant understood how Alexa worked, including the possibility of false wakes and subsequent recordings, it is difficult to see how that person could have a reasonable expectation that communications taking place in the vicinity of the device were private and/or would not be intercepted. Individualized discovery and fact-finding would be necessary to resolve this issue.
vi. Damages
The non-registrant classes seek to recover statutory damages under the relevant state
laws. Plaintiffs argue that because the non-registrants seek “a set amount in statutory damages,”
class wide issues predominate. Dkt. # 257 at 26. While claims for statutory damages sometimes
obviate the need for individualized inquiry, that does not appear to be the case here. California’s
wiretap law authorizes recovery of $5,000 per violation, while the other relevant state laws
provide for liquidated damages at the rate of $100 a day for each day of the violation or $1,000,
whichever is higher. Cal. Penal Code § 637.2(a)(1); Fla. Stat. § 934.10(1)(b); Md. Code Ann.
Cts. & Jud. Proc. § 10-410(a)(1); N.H. Rev. Stat. Ann. § 570-A:11(a); 18 Pa. Cons. Stat.
§ 5725(a)(1); RCW 9.73.060. Plaintiffs offer no method or model to determine the number of
unauthorized recordings made of each non-registrant, making it impossible to ascertain damages
under California law or whether damages are capped under the laws of Florida, Maryland, New
Hampshire, Pennsylvania, and Washington.
See Comcast Corp. v. Behrend
,
vii. Weighing Prevalence and Importance Ultimately, the predominance inquiry requires the Court to determine whether the common issues that lend themselves to representational litigation of the wiretap claims are more prevalent or more important than the issues that will require individualized proof and fact- finding. They are not. Common evidence will likely lead to common answers regarding Amazon’s intent regarding and knowledge of the interception of non-registrant communications, whether Alexa devices are wiretapping “devices” for purposes of the state wiretapping statutes, and whether the universe of publicly-available information regarding how Alexa works was insufficient to establish implied consent for the recordings. While these issues are undoubtedly critical to plaintiffs’ success on the wiretapping claims, they do not come close to establishing either liability or damages. Individual evidence will be needed regarding whether class members expressly consented to the recordings through the acts of their agent or through acceptance of Amazon’s terms and conditions as part of an amazon.com purchase and/or use of the Alexa app, whether class members impliedly consented to the recordings or had a reasonable expectation of privacy around an Alexa device given their individual knowledge and understanding of how Alexa works, and the number of intercepted communications at issue for each non-registrant (or at least whether than number is less than or exceeds ten). Given the number, breadth, and importance of these issues, the Court cannot find that the questions of law or fact common to the non-registrant class members predominate over the questions that will need to be resolved on an individual-by-individual basis. [3]
b. Registrant Class
The registrant class has asserted a single cause of action under Washington’s CPA. To
prevail on a CPA claim, plaintiffs must prove “(1) an unfair or deceptive act or practice, (2)
occurring in trade or commerce, (3) affecting the public interest, (4) injury to a person’s
business or property, and (5) causation.”
Panag v. Farmers Ins. Co. of Wash
.,
i. Injury The proposed class contains tens of millions of people, some of whom defendants assert were never recorded at all, much less had their recordings transcribed, subjected to human review, or used to better defendants’ products. [4] In particular, defendants point out that some registrants use Alexa through a push-to-talk button (eliminating the opportunity for false wakes), may have gifted the device to someone else, or may be non-verbal so that they could not have had their voices recorded. They also argue that almost 900,000 Alexa users have taken advantage of a “do-not-save recordings” function that was introduced in September 2020 and that almost two million Alexa users have opted out of human review. While there is a 1 hypothetical possibility that someone may have purchased and registered an Alexa device but 2 never used it [5] and/or that a non-verbal customer purchased a device advertised as a cloud-based
voice service platform, the Ninth Circuit “do[es] not permit a defendant to support its invocation
of individualized issues with mere speculation.”
Van v. LLR, Inc.
,
With regards to defendants’ other arguments, they each address limited facets of the unfair and deceptive conduct of which plaintiffs complain. If, for example, a registrant uses the push-to-talk feature, there may be less chance of a false wake, but his or her voice is still recorded, transcribed, and subjected to review, storage, and use at defendants’ sole discretion. If a registrant chose the “do-not-save recordings” function when it became available in September 2020, there is no reason to assume that he or she was not recorded prior to that date. And affirmatively opting out of human review does nothing to avoid false wakes, transcription, storage, and use. There being no logical reason to assume, much less evidence to show, that the individual issues defendants identify would bar any class members’ recovery on the CPA claim, the Court will not allow speculation and surmise to tip the decisional scales at class certification. Defendants also argue that an individualized review of every recording associated with a registrant’s account will be necessary in order to determine whether there were any false wakes. But, again, defendants misstate the nature of plaintiffs’ CPA claim. While false wakes are certainly a facet of the device that plaintiffs believe to be unfair, they are also complaining about defendants’ failure to disclose that all recordings – whether intended for Alexa or not -- are transcribed, reviewed, maintained, and used. If plaintiffs prove their allegations regarding how Alexa works and what defendants do with the recordings they collect, the loss of control over their voice data appears to be a class wide injury.
ii. Causation Plaintiffs will have the burden of proving that defendants’ unfair or deceptive conduct ( i.e ., the failure to disclose the prevalence of and/or design preference for false wakes, that recordings are transcribed, that recordings and transcriptions are reviewed by humans, stored, and used indefinitely at defendants’ sole discretion, and that requests to delete recordings left transcripts untouched) caused them to lose the value of and control over their voice data. Defendants argue that each class member’s knowledge regarding how Alexa works and why they purchased the device will have to be evaluated before the fact-finder can determine whether a causal nexus exists. The argument seeks to impose on plaintiffs the task of proving that they would have acted differently in a hypothetical world in which defendants had made adequate disclosures about what they did with the voice data processed by the Alexa service.
Where the act or practice at issue is an omission of material fact, as is the case here, the
causation element of a CPA claim includes a rebuttable presumption of reliance.
Eng v.
Specialized Loan Servicing
,
The presumption can be rebutted in this case by providing evidence that plaintiffs’
decision to purchase and use an Alexa device would have been unaffected had the omitted facts
been disclosed.
G.G. v. Valve Corp
.,
iii. Damages
At the class certification stage, whether the calculation of damages is a common issue
susceptible of class wide determination through common evidence or will require individualized
proof depends on the proposed path for obtaining the necessary data and the method of
computation.
Lytle v. Nutramax Lab’ys, Inc
.,
20 21 there is evidence in the record suggesting that a review of registrant accounts would yield the 22 necessary information. While that review would be individualized,
23 “[t]he amount of damages is invariably an individual question and does not defeat
24 25
class action treatment.” Blackie v. Barrack ,524 F.2d 891 , 905 (9th Cir.1975); see also Yokoyama [ v. Midland Nat’l Life Ins. Co. ,594 F.3d 1087 , 1089 (9th Cir.2010)] (“The potential existence of individualized damage assessments ... does not detract from the action’s suitability for class certification.”). . . . Indeed, the Supreme Court clarified in Dukes that “individualized monetary claims belong in Rule 23(b)(3).” [564 U.S. at 362 ]. Thus, the presence of individualized damages cannot, by itself, defeat class certification under Rule 23(b)(3).
Leyva v. Medline Indus. Inc
.,
iv. Standing Revisiting defendants’ Article III standing arguments from Section II.B. of this Order, the Court finds that the concern that a class of all registrants may contain individuals who were never recorded does not preclude certification. While it is clear that federal courts lack the power to grant relief to a plaintiff who suffered no injury, registrants who were never recorded by Alexa can be identified through the same account review that will establish the amount of damages. While the standing issue will necessitate the same individualized review of Alexa account data as the damage issue, it does not predominate over the common issues in the case.
v. Weighing Prevalence and Importance The Court finds that common issues central to resolution of the registrant class claims predominate over individualized inquiries regarding damages and standing.
2. Superiority
23 In addition to finding that common questions of law or fact predominate with regards to
24 25 the registrant class, the Court must also find “that a class action is superior to other available 26 methods for fairly and efficiently adjudicating the controversy.” Fed. R. Civ. P. 23(b)(3). In 27
28
determining whether a class action is superior, the Court must consider the following list of non- exhaustive factors:
(A) the class members’ interests in individually controlling the prosecution or defense of separate actions; (B) the extent and nature of any litigation concerning the controversy already begun by or against class members; (C) the desirability or undesirability of concentrating the litigation of the claims in the particular forum; and (D) the likely difficulties in managing a class action. Defendants argue that the proposed registrant class would be unmanageable because (i) the class is comprised of tens of millions of people and (ii) defendants have prevailed in every Alexa arbitration claim to reach a decision on the merits. Neither consideration makes class treatment less fair or less efficient than individual litigation in this case. The fact that millions of people were allegedly injured by the same conduct suggests that representative litigation is the only way to both adjudicate related claims and avoid overwhelming the courts. And whether plaintiffs will be able to prove the elements of the CPA claim is not properly before the Court on class certification. It may well be that defendants prevail, in which case they will be released from the claims of all class members. Class treatment is not precluded simply because plaintiffs may not win.
V. C ONCLUSION
For all of the foregoing reasons, IT IS HEREBY ORDERED that: 1. Plaintiffs’ motion to certify a class of registrants under Rule 23(b)(2) to seek
25 injunctive relief under the CPA is GRANTED;
26
27
28
1 2. Plaintiffs’ motion to certify six classes of non-registrants under Rule 23(b)(3) to seek 2 damages under state wiretapping laws is DENIED; and
3
Plaintiffs’ motion to certify a class of registrants under Rule 23(b)(3) to seek damages under the CPA is GRANTED.
DATED this 7th day of July, 2025.
A Robert S. Lasnik
United States District Judge
Notes
[1] This matter can be resolved on the papers submitted. The parties’ requests for oral argument are 27 DENIED. 28
[2]
See Satomi Owners Ass'n v. Satomi, LLC
,
[3] In light of this conclusion, the Court declines to determine whether the possibility that some portion of the non-registrant classes was never recorded by Alexa (and therefore lacks standing) precludes certification.
[4] The CPA claim is also based on allegations that the transcript of a recording was maintained even if the registrant attempted to delete the voice recording and that transcripts, as well as recordings, were stored and used by defendants forever, at their sole discretion.
[5] The only evidence cited for this scenario is the deposition testimony of Jodi Brust. Ms. Brust 26 purchased at least four Alexa devices, two of which she gave away as gifts. Dkt. # 302-2 at 11-15. This testimony in no way suggests that Ms. Brust was not the subject of false wakes, transcription, human 27 review, and/or continued use of her voice data. 28
[6] At the class certification stage, plaintiffs need not prove the actual value of the data taken, they 26 simply have to present a model or method for establishing that value which is viable given the facts of the case. They have done so, regardless whether Jonathan Hochman’s expert opinions stop short of 27 applying the model or method. 28