A proposed class action filed in federal court accuses OpenAI of letting outside contractors read, summarize and grade real ChatGPT conversations without telling users in its terms or privacy policy. The case, Vredenburgh v. OpenAI OpCo, LLC (No. 3:26-cv-10527), was filed September 16 in the U.S. District Court for the Northern District of California, according to Open Class Actions.
The suit came two days after 404 Media published an investigation into the program, which the complaint says is code-named “Project Lily.” Nearly every factual claim in the complaint about how the program works is drawn from that reporting. None of the allegations has been proven, and no class has been certified.
How the review work allegedly runs
According to the complaint, contractors are recruited through a staffing firm, Crossing Hurdles, for jobs with titles like “AI data reviewer” and “chatbot evaluator.” A reviewer picks a task from a dashboard and sees a real user’s prompt, often an entire conversation. The reviewer summarizes what the user seemed to want, reads four ChatGPT responses, scores each from one to seven and writes a rationale that feeds back into model development.
OpenAI runs conversations through an automated “Privacy Filter” before reviewers see them. The complaint notes that the filter’s own documentation calls it a redaction aid, not a guarantee. It also says a reviewer’s screen can show a summary of the user’s past ChatGPT use, which can reveal a name or location, and that some reviewed prompts include users asking ChatGPT to keep what they said confidential.
Where OpenAI disclosed it
The case turns on where OpenAI told users about human review. The complaint says the privacy policy lists eleven kinds of outside companies that receive personal data, and none is a data-labeling or human-evaluation vendor. OpenAI does have a Help Center FAQ that answers “Do humans view my content?” by saying authorized personnel and “trusted service providers” may access content “to improve model performance (unless you have opted out).” The plaintiffs say that page is buried in a collection of about 45 help articles. As a contrast, the complaint points to Google, which shows a notice in its chat interface warning that human reviewers process conversations.
Who would be covered
The proposed class covers everyone in the U.S. who used ChatGPT, free or paid, during the limitations period, with subclasses for California residents and paying subscribers. ChatGPT Enterprise, Business, Team and Edu accounts and API customers are excluded. The complaint pleads eight claims, including violations of California’s Unfair Competition Law, False Advertising Law and Consumer Privacy Act, plus intrusion upon seclusion and unjust enrichment.
Beyond damages and restitution, the plaintiffs want a court order barring OpenAI from sending conversations to outside reviewers without separate opt-in consent, turning the “Improve the model for everyone” setting off by default and adding a warning in the chat window. OpenAI was served September 21, and its response was due October 13. The first case management conference is set for December 18.


WebProNews is an iEntry Publication