1. Overview
The factual scenario at the heart of Bartone and Canu v. Meta Platforms Inc. and Luxottica of America Inc. is as simple as dystopian: a pair of fashionable sunglasses that silently streams the most intimate moments of its wearer’s life to human reviewers overseas.
The case is a federal class action, filed on 4 March 2026, before the United States District Court for the Northern District of California (Case No. 3:26-cv-01897). The action was brought by Clarkson Law Firm on behalf of two plaintiffs – Mateo Canu, resident in Los Angeles, California, and Gina Bartone, resident in New Jersey – who claim to have purchased the Meta Ray-Ban AI smart glasses relying on some privacy assurances that the complaint now seeks to prove as materially false. The proposed class action, therefore, comprehends all US purchasers of the Ray-Ban Meta and Oakley Meta models: a population potentially numbering in the millions, given that the global sales of the device exceeded seven million units in 2025 alone.
The case might therefore raise interesting questions for civil liability of AI wearables: whenever a company places privacy at the core of its products marketing strategy, does its legal exposure remain confined to the contractual perimeter limited by the Terms of Service, or does it attract broader obligations – whether in tort, statutory, or consumer protection law? Furthermore, what is the significance of silence or false claims regarding the protection of consumer privacy?
2. Background: the Swedish investigation that set the case in motion
The complaint draws its factual foundation from a piece of investigative journalism published on 27 February 2026 by the Swedish daily Svenska Dagbladet (SvD), in collaboration with Göteborgs-Posten (GP), about Meta’s data practices for its Ray-Ban AI Glasses (Svenska Dagbladet, 2026).
Although there have been similar investigations, the one published by SvD/GP is the sole journalistic source directly cited in the complaint (complaint, footnotes 6-13, p. 9).
The investigation centres on Sama, a San Francisco-based company with principal operations in Nairobi, Kenya, engaged by Meta as a data annotation subcontractor. Data annotation can be defined as the action of having human workers label raw video, image, and audio data, so that AI systems can be trained to recognise and interpret them. AI systems need to be fed with large amounts of data; proprietary sources of new data, however, can cost hundreds of millions of dollars (complaint, footnote 14, Hamidah Oderinwale and Anna Kazlauskas, The Economics of AI Training Data: A research Agenda, April 2026). Meta seems able to farm personal data directly from the pool of consumers who have purchased the AI glasses, streaming footage captured through the Ray-Ban AI to Sama’s facilities in Nairobi, where workers manually review and tag the content to train Meta’s AI models. The conditions under which that review took place are central to the complaint’s factual narrative; the reporters found out that human annotators were systematically exposed to footage of deeply intimate nature: individuals undressing, using the bathroom, engaging in sexual activities, handling sensitive financial documents – all (inadvertently?) captured by the glasses, worn in everyday domestic settings. For example, the complaint reports the account of a worker who was asked to annotate the footage of a woman who unknowingly undressed in front of her partner’s glasses left on a bedside table (complaint, par. 42, p. 9). Workers also reported that Meta’s automated facial-blurring system – the main technical privacy safeguard featured in the company’s consumer communications – frequently failed, leaving faces visible in footage streamed for the annotation. In addition to the American class action under consideration, the same investigation prompted a formal inquiry by the UK Information Commissioner’s Office and a parliamentary question to the European Commission (Parliamentary Question P-000903/2026).
3. Core Allegations: False Promises, No Opt-Out, Disclosure Inadequacy
It’s worth noting that, despite the highly sensitive character of the data at issue, the complaint’s focus is not on the infringement of the right to privacy, but on a consumer protection action, organised around three different allegations: false advertising, absence of any opt-out mechanism, and inadequate disclosure.
3.1 False Advertising
The plaintiffs allege that Meta – and Luxottica as co-manufacturer and co-advertiser – mislead consumers to purchase the Ray-Ban AI glasses through representations that were materially false; the marketing language at issue appeared on Meta’s AI glasses dedicated marketing page, and included a main claim « Designed for privacy, controlled by you » (complaint, par. 37, p. 7), along with other sentences such as « You’re in control of your data and content »; « Clear, easy device and app settings help you manage your information, giving you control over what content you choose to share with others, and when » (par. 38, p. 8), and «Built for your privacy and others’ too» (par. 39, p. 8). Clarkson collectively labels these statements « Challenged Representations and Omissions »; and asserts that their falsity is grounded on the fact that product’s core AI functionality necessarily requires the transmission of footage to Meta’s servers – and, consequently, to annotators overseas – without any control by the user over the stream of data.
3.2 Absence of an Opt-Out Mechanism and consequent Loss of Value
The second allegation concerns the fact that, because the glasses must be connected to activate any AI functionality, the use of the AI assistant is technically inseparable from data transmission – and, consequently, human review. Any consumer unwilling to participate in that data stream is «left with a $299 to $799 pair of sunglass frames, with no AI or smart functionality whatsoever» (complaint, par. 53, p. 12).
3.3 Terms of Service and Disclosure Inadequacy
Lastly, the third allegation argues that Meta’s Terms of Service provision on human review is too vague to constitute a meaningful informed consent. In fact, a careful reading of the Supplemental Meta Platforms Technologies Privacy Policy, Section IV (« How do we process information when you use our voice services for Meta Wearable Products? »), reveals that « The Meta AI service on AI Glasses (if available for your device) stores text transcripts and audio recordings of your voice interactions by default to help improve Meta’s products (…) We use machine learning and trained reviewers to process this data to improve, troubleshoot, and train our products. We share that information with third party vendors and service providers to improve our products. You can access and delete recordings and related transcripts in the Meta AI app. »
Three observations follow: first, the provision seems to be confined to voice services, leaving unaddressed the processing of images and video data; second, the terms contain no explicit passage warning users against sharing sensitive persona information with the AI, nor any indication that such information may be used to train AI systems; third, the document is silent on data retention period, volume of data subject to review and access to that data. The sole mechanism mentioned is user-initiated deletion.
In conclusion, the complaint states: « By highlighting the Meta AI Glasses’ privacy features, while concealing the critical fact that intimate footage is routinely reviewed by Defendants’ employees and contractors, Defendants created a false impression that the Glasses provided comprehensive protection of the user privacy, when in reality the marketed privacy features (…) do nothing to prevent the most significant privacy risk the Glasses pose. » (complaint, par. 52, p. 11).
4. Legal Claims and Relief Sought
The charges brought by the plaintiff can be summarized as violations of rules pertaining to three distinct legal frameworks.
The first and broadest framework is the California consumer protection law. Under the California Unfair Competition Law (UCL, Cal. Bus. & Prof. Code §§ 17200 et seq.), the complaint raises all three statutory prongs – unfair, fraudulent, and unlawful business practices – against Meta’s privacy marketing campaign (complaint, par. 80-118, pp. 19-25). The California False Advertising Law (FAL, Cal. Bus. & Prof. Code §§ 17500 et seq.) adds a prohibition on untrue or misleading advertising (par. 119-130, pp. 25-27), while the California Consumers Legal Remedies Act (CLRA, Cal. Civ. Code §§ 1750 et seq., specifically §§ 1770(a)(5), (7), (9)) prohibits misrepresenting a product’s characteristics or quality and advertising with intent not to sell as described (par. 131-142, pp. 27-29). For Bartone, resident in New Jersey, the New Jersey Consumer Fraud Act (NJCFA, N.J. Stat. §§ 56:8-1 et seq.) extends similar consumer fraud logic to her home jurisdiction (par. 143-152 pp. 29-30).
The second framework is common-law fraud, which the complaint distinguishes into two different aspects: misrepresentation and concealment. The count of fraud by misrepresentation targets the affirmative statements in Meta’s marketing campaign, as positive assertions of facts known to be false at the time of sale (par. 153-160, pp. 30-31). The count of fraud by concealment instead targets the deliberate omission of the Sama annotation pipeline as a fact that an average consumer would have considered in a decision of purchasing or not the product (par. 161-172, pp. 31-33).
The third framework is grounded in contract: a breach of contract derives from the fact that Meta’s privacy representations formed an express term of sales, the non-performance of which entitles the customer to compensation. (par. 180-185, pp. 33-34).
Regarding the relief sought, instead, the plaintiff pursues three different categories of remedy.
The first is injunctive and corrective relief, seeking a court order that requires the defendants to cease marketing, advertising, distributing, and selling the Ray-Ban AI glasses in a manner alleged to be unlawful. However, the object is not merely to stop the conduct, but also to address the informational asymmetry created in the market, by undertaking an « affirmative advertising campaign to correct the public misleading impressions of the products resulting from defendants’ unlawful conduct » (injunction, p.37).
The second is monetary relief, sought in the forms of «damages, restitution and/or disgorgement ». This comprehends losses suffered by individual purchasers, return of any benefit unfairly obtained by defendants, and disgorgement of profits derived from the misleading campaign.
The third is punitive damages, based on the fraud and knowing-violation counts. Under US law, punitive damages are applicable whenever the behaviour is clearly reprehensible, aiming to discourage similar conduct in the future. These can arise from statutory violations, breach of contract, fraud, bad faith etc.; the claimant’s request, therefore, makes it evident that this is not simply a commercial or compensation dispute but rather a full-scale legal battle, aimed at rectifying certain deplorable practices carried out by Big Tech companies.
5. Concluding Remarks
As of May 2026, Meta has filed no formal defence yet in Bartone and Canu v. Meta Platforms Inc. and Luxottica of America Inc.; the next procedural step is the initial case management conference, scheduled for 5 June 2026. In the meantime, the litigation has already expanded: Clarkson Law Firm uploaded a first amended complaint, increasing the number of named plaintiffs from two to nineteen and extending the proposed subclasses from two to sixteen US states (Case No. 3:26-cv-01897-EMC). At the same time, it looks like Meta has reviewed its relationship with Sama, giving termination to the contract.
In a more domestic perspective, if a materially identical case were brought before European Courts, this would engage in a significantly more prescriptive regulatory framework. The Unfair Commercial Practices Directive (EC/29/2005) provides a background that is analogous to UCL and FAL claims, prohibiting misleading commercial actions (Article 6) and misleading omissions (Article 7). On the other hand, the GDPR (Regulation 679/2016) would add extra layers of protection, regulating purpose limitation and data minimisation (Article 5), obligations to identify processors and data recipients at the point of collection (Articles 13-14), and conditions to transfer data to third countries (Articles 44-49). For example, it would be interesting to find out whether Sama’s operation in Kenya meets the data processing quality requirements set out in the GDPR. Finally, the EU AI Act (Regulation 1689/2024) qualifies high-risk AI systems, triggering requirements of transparency and prior conformity assessment. These concerns have already surfaced at the highest level of EU institutions, in the Parliamentary Question mentioned above. Whether the Commission’s response will translate into any action against Meta remains to be seen, but what the Bartone and Canu v. Meta Platforms Inc. and Luxottica of America Inc case has already made evident on both sides of the Atlantic is that, for companies, the cost of remaining silent – or worse, making false statements – regarding data processing can no longer be easily contained.
