OpenAI faces lawsuit over alleged human review of ChatGPT chats
Two California users allege contractors reviewed ChatGPT conversations without clear disclosure, seeking consent rules and damages.
By Dev Ramirez · Crypto Correspondent
· 3 min read
A ChatGPT human review lawsuit filed by two California users alleges that OpenAI allowed outside contractors to read and assess real conversations without clearly telling consumers. The proposed class action has not produced a court finding of wrongdoing, and the available reports do not include OpenAI’s response.
The case, Vredenburgh, et al. v. OpenAI OpCo LLC, No. 3:26-cv-10527, was filed in the U.S. District Court for the Northern District of California, according to Top Class Actions. For people using ChatGPT for personal, work or financial questions, the case centers on who may see the text they enter and what disclosures users received.
What does the ChatGPT human review lawsuit allege?
The plaintiffs say an internal program called Project Lily sent real user prompts, and in some cases full conversations, to third-party contractors working in roles such as AI data reviewers or chatbot evaluators, Decrypt reported. The contractors allegedly summarized a user’s goal and evaluated several possible chatbot replies, assigning scores on a one-to-seven scale.
That kind of evaluation is one way AI developers can assess whether a model’s answers are useful or appropriate. The suit’s central claim is narrower: plaintiffs contend OpenAI did not adequately disclose that outside human reviewers could read the substance of conversations for model-improvement work, according to Decrypt and Top Class Actions.
The complaint also alleges that OpenAI applied an automated privacy filter before chats reached reviewers, but that personal information could still remain in the material shown to contractors. Top Class Actions reported that the plaintiffs say OpenAI’s consumer materials disclosed some ways content could be accessed, including safety reviews, but did not clearly identify third-party human evaluators reviewing chats for model development.
What are the plaintiffs asking OpenAI to change?
The proposed class action asserts California consumer and privacy claims, along with common-law privacy-related claims including intrusion upon seclusion. Those are allegations, not findings of liability. Top Class Actions reported that the complaint also cites California’s Unfair Competition Law, Consumer Privacy Act, Consumers Legal Remedies Act and False Advertising Law.
The plaintiffs seek damages, restitution and other monetary relief, plus court orders that would change the alleged practice. According to Decrypt, their requested injunctions include requiring opt-in consent before a chat is sent to an outside reviewer, making the “Improve the model for everyone” setting off by default, and adding a clear notice in the chat interface. The request also contemplates deleting work tied to reviewed conversations and retraining models that used it.
Decrypt reported that OpenAI was served on Sept. 2 and had until Oct. 13 to respond. The evidence available for this report does not establish whether the company filed a response, how broadly any contractor review occurred, or whether a court will accept the plaintiffs’ claims.
This story draws on original reporting from Decrypt.
