A new federal lawsuit alleges that OpenAI routes real ChatGPT prompts and conversations to outside contractors who read, summarize, and score them to improve models, without disclosing this in its Terms of Use or Privacy Policy. The case, _Vredenburgh v. OpenAI Op Co, LLC_, No. 3:26-cv-10527, was filed on September 16, 2026 in the U.S. District Court for the Northern District of California in San Francisco.

The complaint, citing a 404 Media investigation, describes a review pipeline code-named “Project Lily.” Contractors hired through staffing firm Crossing Hurdles work from a dashboard showing real user prompts and conversations. They write summaries, score responses from one to seven, and provide rationales. The complaint alleges personal details can bypass an automated “Privacy Filter,” which is described as a redaction aid, not a guarantee.

The case centers on a gap between what OpenAI told users and where it disclosed practices. The Privacy Policy lists eleven types of outside vendors receiving personal data but omits data-labeling or human-evaluation vendors. The model-training page mentions automated removal of personal information and machine training, with no mention of human readers. The only disclosure of human review limits it to flagged content checked by “our team” for policy violations.

OpenAI has not yet responded in court, and none of the allegations has been proven. The proposed class includes everyone in the U.S. who used ChatGPT Plus, with California and paid-subscriber subclasses, but excludes Enterprise, Business, Team, Edu, and API users. There is nothing to claim at this time.