- Lead. A coalition of 42 state attorneys general, led by New York Attorney General Letitia James, served OpenAI with a formal subpoena on June 12 — just four days after the company filed its S-1 registration statement with the SEC, creating an immediate material-risk disclosure obligation for what could be one of the largest IPOs in US history.
- Fact. The subpoena demands records covering advertising and marketing practices, health data collection, treatment of minors and seniors, internal safety testing policies before product releases, and a specific AI design property called “model sycophancy” — the documented tendency of large language models to tell users what they want to hear rather than what is accurate.
- Stake. The investigation represents the broadest state-level legal action yet taken against an AI company, and signals that US states intend to use existing consumer protection law against AI developers without waiting for congressional or federal action.
OpenAI had filed its S-1 confidentially with the Securities and Exchange Commission on June 8, with Goldman Sachs and Morgan Stanley as underwriters and a reported valuation between $850 billion and $1 trillion. The New York subpoena arrived on June 12, forcing OpenAI to disclose the multistate investigation as a material legal risk before its planned listing — reportedly targeting September 2026 — can proceed, according to TechCrunch’s reporting on the investigation.
What the Subpoena Covers
The scope of the demands is notably specific. Beyond standard advertising and data practices, the subpoena expressly names model sycophancy as a focal point — a documented byproduct of reinforcement learning from human feedback (RLHF), the training method that has underpinned ChatGPT’s development. The investigation also targets OpenAI’s internal safety testing policies and procedures prior to product releases, an area where regulators have struggled to obtain meaningful transparency from AI developers.
OpenAI said it takes the concerns “seriously” and intends to “engage constructively” with the attorneys general, pointing to recent product changes including “a more protective experience for minors and people experiencing difficult situations.” Florida is running a parallel criminal investigation into OpenAI’s potential role in the April 2025 Florida State University mass shooting, where the alleged perpetrator exchanged more than 13,000 messages with ChatGPT, including queries about campus schedules minutes before the attack.
Consumer Law as the AI Accountability Tool
The multistate approach reflects a choice by US states to apply existing consumer protection law to AI, rather than waiting for federal AI-specific legislation. The strategy differs markedly from the European approach: the EU has backed the development of open-source public AI infrastructure as its primary governance tool, while simultaneously advancing sector-specific regulation. US states, by contrast, are treating AI companies as consumer-facing product manufacturers and applying the liability framework accordingly.