← Back to feed

A coalition of 42 state attorneys general has opened an investigation into OpenAI. New York’s attorney general served the company with a subpoena on June 13, 2026, five days after OpenAI filed confidentially for an IPO at an $852 billion valuation. The Wall Street Journal reported the investigation first; OpenAI confirmed it the same day.

The subpoena requests documents on a broad range of topics: advertising practices, user engagement and retention, model sycophancy, consumer and health data handling, treatment of minors and seniors, and internal company policies. OpenAI declined to name which states are participating or to provide further detail on what was requested. The company said it “takes the concerns raised by state attorneys general seriously” and intends to “engage constructively.”

IPO Timing

OpenAI filed confidentially for its IPO on June 8. Goldman Sachs, Morgan Stanley, and JPMorgan are leading the offering. The company closed a $122 billion funding round in March at the $852 billion valuation that underpins the listing.

A multistate investigation of this scale constitutes a material legal proceeding that must be disclosed in the company’s S-1 prospectus. That creates a direct complication: the investigation is now a risk factor in a public filing that will be scrutinised by institutional investors, underwriters, and regulators before trading begins. Anthropic also filed confidentially for an IPO last week at a $965 billion valuation, making this a crowded AI IPO window with competing disclosures.

What the Subpoena Covers

The scope of New York’s subpoena suggests an inquiry into whether OpenAI’s business practices and product design caused harm to users, particularly vulnerable populations. The listed topics break into three areas:

Consumer protection: Advertising practices, user engagement and retention mechanics, how ChatGPT is marketed to the public.

Data and safety: Handling of consumer data and health data, treatment of minors and seniors, internal safety testing policies before product releases.

Model behaviour: The inclusion of “model sycophancy” as an explicit topic is notable. It treats AI output quality — specifically, models that tell users what they want to hear rather than what is accurate — as a potential consumer protection violation rather than a purely technical issue.

The Legal Escalation

The multistate probe follows a rapidly escalating sequence of legal actions against OpenAI:

• June 1: Florida became the first US state to sue OpenAI. Attorney General James Uthmeier filed an 83-page complaint naming CEO Sam Altman personally, treating ChatGPT as a defective product under product liability law. The Florida complaint focuses on a shooting where the suspect had used ChatGPT.
• June 8: OpenAI filed its IPO confidentially with the SEC.
• June 13: New York served the multistate subpoena. Florida’s AG had publicly said he expected other states to follow; 42 apparently did.

Separately, a criminal investigation into OpenAI is underway in at least one jurisdiction. OpenAI also faces active copyright litigation from publishers and authors over Llama training data, and an ongoing civil suit from co-founder Elon Musk (currently on appeal after Musk lost on statute of limitations grounds).

Consumer Protection as AI Regulation

The multistate investigation is structurally different from the regulatory approaches seen in Europe. EU AI Act enforcement targets categories of risk based on use-case classification. The US state-level approach being deployed here is consumer protection law — the same toolkit used against deceptive advertising, data brokers, and defective products.

That framing has practical consequences. Consumer protection cases can move faster than federal rulemaking. They can seek injunctions, civil penalties, and changes to product behaviour. The model sycophancy line in the subpoena suggests that state enforcers are willing to treat AI product design decisions as actionable harms, not just theoretical risks.

OpenAI’s congressional testimony — from Sam Altman and others — has consistently framed regulation as something OpenAI welcomes in principle. The gap between that framing and 42 simultaneous state investigations is now part of the company’s prospectus.