Meta Glasses plaintiffs extend their training-data claim to people who never wore the glasses

A capture LED and a delete button both sit on the buyer's side of the transaction: Meta's product website describes an indicator that lights when the camera records, and software through which users manage or delete photos, videos and voice interactions ^[5]. That is the design fact the amendment goes at. Notice that a camera is on is a different object from disclosure that the resulting frames go to Meta's servers for analysis and, the plaintiffs allege, into model training ^[16], and a person walking past the wearer has no account, no setting and no point in that pipeline at which a preference could be registered.

Try to price the exposure and the arithmetic runs out quickly, which is the honest finding. The only figure in the record is the child-safety settlement Meta agreed to, up to $18 billion over the next decade ^[1], which averages about $1.8 billion a year ^[14], and that is a separate matter with a separate class; nothing here attaches a number, a statute or a class size to the glasses case. What the complaint does supply is an allocation fact, or rather the more interesting version of one: on the plaintiffs' account, Meta made no effort to disclose the practices to bystanders or seek their consent ^[4], meaning the consent budget was spent entirely on the purchaser.

The evidentiary spine is the February 2026 reporting by Svenska Dagbladet and Goteborgs-Posten, drawing in part on interviews with annotation workers at Sama, the Kenya-based outsourcer that did data-annotation work for Meta ^[6], including the allegation that some workers saw identifiable faces despite Meta's purported anonymization ^[9]. Ryan Clarkson of Clarkson Law Firm, co-lead counsel ^[11], describes the incentive as corporations hunting more signal and more recordings to feed into the machine ^[12], and the hardware and software together as a surveillance economy ^[13]. Meta says it disagrees and will fight ^[10].

The disagreement over what happens next splits three ways, diverging on mechanics rather than sentiment. The first: the court requires each bystander to show they appear in identified footage, individual proof swamps the class, and the case narrows back to people who paid for a product marketed as designed for privacy and controlled by you ^[2]. The second: the bystander theory survives pleading, at which point the cost of shipping a camera with an AI mode ^[17] includes a capture-time disclosure that no glasses maker currently offers. The third: the LED plus the in-app controls are held to be adequate notice, and the surviving claim is a false-advertising claim about a marketing line.

The second is the one worth pricing, for an unglamorous reason: designing a bystander-facing disclosure is cheaper than litigating whether one was owed, and the party that builds it first sets the standard the others get compared against. The evidence does not carry more than that. It contains no ruling, no certification decision and no damages theory, so anyone quoting a number for third-party liability here is quoting their own model. What would break the thesis is a dismissal of the bystander claims on standing or manageability before discovery, or a showing that the human-review pipeline was disclosed somewhere a person who never bought the glasses could plausibly have read it.