OpenAI is facing a multi-state investigation as attorneys general examine the company's data practices, safety measures, and the potential impact of its artificial intelligence products on consumers ahead of its anticipated initial public offering. A coalition of state attorneys general has launched the investigation and served the company with a subpoena, according to a Wall Street Journal report citing people familiar with the matter.

Subpoena Details and Scope

The subpoena, reportedly issued by New York Attorney General Letitia James’ office, seeks documents related to a broad range of operational topics. The request covers advertising practices, user engagement and retention strategies, and consumer and health data handling. Investigators are also requesting information on activities involving minors and seniors, deep-learning models, AI sycophancy, and internal company policies. AI sycophancy refers to situations in which chatbots excessively agree with or reinforce users’ views rather than providing balanced responses.

Regulatory and Legal Context

The reported investigation comes shortly after OpenAI confidentially filed paperwork with the Securities and Exchange Commission for a potential IPO. The company is also facing legal scrutiny elsewhere. Florida became the first state to sue OpenAI and CEO Sam Altman earlier this month, alleging they knowingly released an unsafe product despite warnings about potential risks. In April, Florida Attorney General James Uthmeier opened a criminal investigation into OpenAI over ChatGPT’s alleged role in assisting a suspect involved in a mass shooting at Florida State University.

Broader Industry Scrutiny

OpenAI is not alone in attracting attention from state regulators. In December, a coalition of 42 attorneys general sent a letter to OpenAI, Meta Platforms, Inc., Anthropic, Alphabet Inc.'s Google, and SpaceX's xAI. The letter urged stronger safeguards for vulnerable users and warned developers could face accountability for harmful chatbot outputs. California Attorney General Rob Bonta also announced an investigation earlier this year into sexually explicit images allegedly generated using xAI’s Grok chatbot.

For the past few years, public market exposure to artificial intelligence has largely meant buying a handful of companies that sit underneath the models themselves. Microsoft was the primary conduit to OpenAI, investing more than $13 billion and embedding its systems across Azure and Office. Alphabet and Amazon have taken a similar role with Anthropic, backing the company with multi-billion-dollar commitments while integrating its models into their cloud ecosystems. And in the middle of the compute layer sits Nvidia, which has a data center segment that has become increasingly tied to demand from AI training and inference workloads. Together, these firms have defined what investors think of as the "AI trade," not ownership of the models, but ownership of the systems that make them possible.

That structure may be starting to shift. OpenAI and Anthropic have both filed confidential preparations for eventual public listings, while SpaceX is set to IPO this month. This current IPO pipeline is set to mark one of the most significant transitions of the AI era: moving from indirect exposure through hyperscalers to direct ownership of frontier model developers and adjacent infrastructure platforms.

The scale of those private companies underscores the potential impact. Anthropic has recently been valued at $965 billion, reflecting intense investor demand for leading foundation model providers. The company is also believed to be operating at an annualized revenue run rate approaching $50 billion, driven by rapid enterprise adoption of its Claude models, though those revenues come with substantial compute and training costs that remain opaque to public investors, CNBC reported. OpenAI, meanwhile, is generating billions in annualized revenue and reports that its products now reach hundreds of millions of weekly users, highlighting the speed at which consumer and enterprise AI adoption has scaled before any public-market scrutiny, Reuters reported.

SpaceX, while outside pure AI software, has become part of the broader frontier-technology capital stack. The Globe and Mail reported SpaceX believes its largest growth opportunity is artificial intelligence. The company has reportedly been valued in private markets in the $1.5 trillion to $2 trillion range in secondary transactions and fundraising discussions, reflecting the scale of demand for vertically integrated space, communications, and defense-adjacent infrastructure platforms.

Taken together, these listings could bring several trillion dollars of new equity to public markets over time—large enough to reshape the makeup of the global mega-cap technology cohort. But the more consequential shift may not be their size. Public markets would force a level of financial disclosure that private funding rounds have largely obscured. Investors would be able to assess how effectively leading AI companies convert massive compute expenditure into durable revenue, how dependent they remain on hyperscaler distribution, and whether current growth rates justify valuations built on rapidly evolving technology cycles.

It would also expose the circular nature of the AI ecosystem more directly. Hyperscalers fund model developers, model developers drive demand for compute, and compute providers feed back into hyperscaler revenue growth. In private markets, that loop has reinforced valuations across the stack. In public markets, it would be subject to quarterly scrutiny. When these companies do list, they will not simply expand the AI trade but redefine it. The question for investors will no longer be which layer of the stack benefits most from AI adoption. It would be whether the companies building the models can sustain the economics of the infrastructure they helped create.