Coalition of 42 state attorneys general opens investigation into OpenAI, subpoenas documents on data practices and model sycophancy

A coalition of 42 U.S. state attorneys general has opened a sweeping investigation into OpenAI, with New York's attorney general leading the effort by serving the company with a broad subpoena. The probe, which emerged publicly on June 12, 2026, covers a wide range of OpenAI's activities — from how it handles consumer and health data to how its models engage with minors, seniors, and vulnerable users.

What's new

The subpoena from New York's AG seeks documents across several categories:

• Advertising practices and user engagement and retention strategies
• Handling of consumer and health data
• Activities and protections related to minors and seniors
• Deep learning model documentation
• Internal company policies
• Information on model sycophancy — the tendency of AI models to tell users what they want to hear rather than what is accurate

OpenAI said it is cooperating with the probe. A company spokesperson said: "AI is a new and powerful technology, and we work every day to safely bring its benefits to people in a responsible way."

Context

The 42-state coalition represents a coordinated regulatory effort that goes well beyond any single state's jurisdiction. New York's AG is coordinating the group, which marks the broadest state-level legal action against an AI company to date.

OpenAI has faced a growing legal docket in 2026. Florida filed its own lawsuit earlier this year, alleging the company misrepresented the safety of ChatGPT in connection with a mass shooting. The state AG investigation is distinct from and broader than the Florida action.

OpenAI is preparing for an IPO expected in September 2026. The investigation's timing — months before the listing window — means any material findings or enforcement actions would need to be disclosed to prospective public investors.

Why it matters

The inclusion of model sycophancy as an explicit investigation subject is notable. It signals that regulators are moving beyond data privacy and platform safety concerns toward scrutiny of the models' behavioral properties — how AI systems influence what users believe and decide. That is a harder problem to regulate and a harder one to remedy with a policy change.

A 42-state coalition also has the practical weight to demand broad document production that a single state could not. If the investigation surfaces systematic findings across advertising, retention, and data practices, it could shape product decisions and disclosure requirements in ways that extend well beyond this specific probe.

For OpenAI's IPO, the investigation adds regulatory overhang. Public investors will need a clear view of the probe's scope and potential liability before the listing window opens in September.