A privacy lawsuit has to answer a question the privacy harm itself does not: how much money? The class action filed against OpenAI on September 16, 2026, Vredenburgh v. OpenAI OpCo, LLC, alleges that contractors read real ChatGPT conversations to grade the chatbot under a program called “Project Lily,” without telling users. The complaint asks for damages, restitution and punitive damages on behalf of a proposed class covering every U.S. ChatGPT user. OpenAI has not yet responded and nothing has been proven. There is no settlement and nothing to claim. But the damages theories are already on paper, and they show how lawyers try to put a number on a read conversation.
Theory One: The Subscriber Price Premium
Paying users — ChatGPT Plus, Pro and other consumer subscribers — get their own subclass. Their theory is that they paid for a service whose privacy was misdescribed, and that some part of what they paid was for privacy they did not receive.
This is the most familiar damages model in consumer class actions. It does not require proof that any individual’s chat was read. It requires an estimate of how much less the product would have been worth, or how much less consumers would have paid, if the truth had been disclosed. Plaintiffs typically support that figure with an expert survey, often a conjoint analysis, that measures what buyers will pay for a feature. The difficulty is that it must isolate privacy from everything else a subscription buys: faster models, higher usage limits, newer features.
Theory Two: The Value of a Prompt
Free users paid nothing, so a price premium does not work for them. The complaint’s answer is that every user gave OpenAI something of value: their prompts. It calls real user prompts some of the scarcest raw material in the AI industry and points to marketplaces where prompts are bought and sold to argue that the value is measurable.
This theory matters because of standing. California’s Unfair Competition Law and False Advertising Law require a plaintiff to have lost money or property. Courts have split on whether personal information counts. If the prompt-value theory holds, free users have a path to restitution; if it fails, the free-user portion of the class may be limited to the privacy torts.
Theory Three: Unjust Enrichment and Restitution
Restitution looks at the defendant’s gain rather than the plaintiff’s loss. The complaint’s unjust enrichment claim targets whatever value OpenAI extracted from the reviewed conversations — the improvement in its models that the contractors’ ratings allegedly produced. Measuring that would be contested and expert-heavy: how much of a model’s quality is attributable to one pipeline of preference data among many?
Theory Four: Statutory Damages Under the CCPA
The California Consumer Privacy Act allows statutory damages of $100 to $750 per consumer per incident, or actual damages if greater, with the amounts now adjusted for inflation by the state. Statutory damages are attractive in a huge class because they do not require proof of individual harm. The catch is that the CCPA’s private right of action covers security failures that lead to unauthorized access or disclosure. The complaint frames OpenAI’s conduct as sharing personal information without reasonable safeguards; whether the claim survives in that form is a threshold question. If it does, the California subclass carries a potentially large statutory number. That is a large part of why it exists as a separate subclass.
Theory Five: Punitive Damages
The complaint seeks punitive damages, tied mainly to its fraudulent omission and concealment claim. In California, punitive damages require clear and convincing evidence of oppression, fraud or malice. The plaintiffs’ strongest facts for that standard are the ones suggesting OpenAI knew how to disclose human access — it warns users when a workspace administrator can read their chats — and chose not to for model training, and the complaint’s allegation that OpenAI did not answer 404 Media’s question about where users had been told until after the story ran.
The CLRA Timeline
The Consumers Legal Remedies Act claim currently seeks only an injunction. That is required: the statute makes a plaintiff give the business 30 days’ notice before suing for damages. The plaintiffs sent OpenAI a demand letter on September 16 and say they will add CLRA damages if OpenAI does not respond within 30 days. If that happens, the change would come through an amended complaint.
The Relief That Is Not Money
For many readers the injunctive demands matter more than the damages. The complaint asks for separate opt-in consent before any conversation goes to outside reviewers, the “Improve the model for everyone” setting switched off by default, a warning in the chat window, and an order requiring OpenAI to delete the reviewers’ work product and stop using or retrain any model built from it. That last remedy would cost far more than most damages figures, and it gives OpenAI a strong reason to fight rather than settle quickly.
What a Class Member Could Realistically Expect
Nothing, yet. The case is at the complaint stage: OpenAI’s response is due October 13, 2026, and the first case management conference is December 18. Mass privacy settlements that cover tens of millions of people usually pay individual amounts that are small relative to the headline, because the fund is split across everyone who files. For a sense of how privacy claims get valued once real numbers exist, see what the LifeMD telehealth privacy claim was actually worth and how much you can sue for in a data breach. The complaint and a live case summary are at OpenClassActions.com, with class and date tracking at OpenClassActions.org.
Frequently Asked Questions
How much money could ChatGPT users get from the Project Lily lawsuit?
There is no figure. The case was filed on September 16, 2026, OpenAI has not responded, no class is certified and there is no settlement. The complaint seeks damages, restitution and punitive damages but does not state a per-user amount.
Why do paying ChatGPT subscribers have a separate subclass?
Because their damages theory is different. Subscribers allege they paid a premium for a service whose privacy was misdescribed. Free users rely on a theory that their prompts had value OpenAI took without disclosure.
Can free ChatGPT users recover money?
Possibly, but their path is harder. California’s UCL and FAL require a plaintiff to have lost money or property, and courts disagree on whether personal data counts. The privacy tort claims do not have the same requirement.
What are CCPA statutory damages?
Between $100 and $750 per consumer per incident, or actual damages if greater, with the amounts adjusted for inflation. They apply to the CCPA’s private right of action, which covers security failures; whether this case fits that category is disputed.
Why is the CLRA claim injunction-only?
The Consumers Legal Remedies Act requires 30 days’ written notice before a damages claim. The plaintiffs sent notice on September 16, 2026 and say they will add damages if OpenAI does not respond.
Sources
- Class Action Complaint, Vredenburgh v. OpenAI OpCo, LLC, No. 3:26-cv-10527-AGT (N.D. Cal., filed September 16, 2026).
- Court docket on CourtListener: service on September 21, 2026, the October 13 response deadline and the case management schedule.
- 404 Media, “Inside Project Lily: The Humans Reading Your ChatGPT Chats” (September 14, 2026), the reporting the complaint relies on.
- OpenAI Help Center, “Data Usage for Consumer Services FAQ”; OpenAI Privacy Policy and “How your data is used to improve model performance”.
- OpenClassActions.com: Humans Are Reading Your ChatGPT Chats, New Class Action Claims — case summary, complaint PDF and status tracking.
This article is for informational purposes only and is not legal advice. LawyerDamages.com is not a law firm and is not affiliated with OpenAI or any party to the case. Everything described here is an allegation in a complaint: OpenAI has not yet responded in court, no class has been certified and no court has decided the merits. There is no settlement and nothing to claim. Anyone asking for a fee to “join” this case is not part of it.