Skip to main content
Bureau of AI
The state of ai impact assessment
N: Legal or test findingLegal actionAlleged

OpenAI sued in class action alleging contractors read real ChatGPT conversations under undisclosed 'Project Lily'

On September 16, 2026 a proposed class action, Vredenburgh v. OpenAI OpCo, LLC, was filed in the U.S. District Court for the Northern District of California. It alleges that OpenAI routed real ChatGPT prompts and conversations to outside contractors, recruited through the staffing firm Crossing Hurdles, who read, summarized and scored them for model training without adequate disclosure to users. The complaint, which draws on 404 Media reporting, cites more than 900 million ChatGPT users worldwide and pleads eight causes of action, including privacy and consumer protection claims. It seeks damages and injunctive relief. The allegations are unproven and OpenAI has not yet responded in court.

Happened September 16, 2026 · Disclosed September 16, 2026

ChatGPT users worldwide cited in complaint
More than 900 million

Who is exposed

Consumer ChatGPT users in the U.S. (free and paid) whose conversations may have been shown to third-party human reviewers. Enterprise, Business, Team, Edu and API customers are excluded from the proposed class.

What to do

Check the 'Improve the model for everyone' data-control setting on any consumer ChatGPT accounts used for work, and steer sensitive work to enterprise or API tiers covered by business agreements. Review vendor privacy disclosures for human-review practices.

Rules it touches

California consumer protection laws (UCL, CLRA, FAL), the California Consumer Privacy Act, the California constitutional privacy right, and common-law intrusion upon seclusion, fraudulent concealment and unjust enrichment claims.

Harm, as stated in the sources

Who was involved

As named in the sources. Parties are alleged unless a source reports a finding or an admission.

Other facts

Application Security Isn’t Optional Anymore.