What State Officials Are Investigating About OpenAI

State attorneys general across the country are looking into OpenAI on two main concerns: whether the company is protecting children who use its services, and questions about how the company restructured its ownership, according to a Bloomberg report from June 13, 2026.

This scrutiny has been building for months. In August 2025, the attorney general of Arizona joined 44 other state attorneys general in sending a formal warning letter to major AI companies about design practices that could harm children. That same month, the attorneys general of California and Delaware singled out OpenAI in their own letter, raising child safety concerns.

Two Separate Tracks

The child safety concerns and the questions about ownership structure have moved forward separately, though they involve some of the same officials.

On the ownership question: OpenAI changed its legal structure in 2025. This matters because OpenAI started as a nonprofit organization — think of it like a charity — but was moving toward a different setup. Since California and Delaware are the states where OpenAI was officially registered, their attorneys general reviewed this change. California's AG issued a statement about it in October 2025. Delaware's AG announced the same month that her office had finished its review and said the process led to safety improvements and other protections.

The child safety track moved in a different direction in late 2025. In North Carolina, two officials launched a nationwide task force in November 2025 to focus on AI safety issues. OpenAI and Microsoft agreed to take part directly — a more collaborative approach than a formal investigation, though the two can happen at the same time.

This collaborative effort produced results. In April 2026, OpenAI released a Child Safety Blueprint that was developed with the North Carolina officials. The document explains how OpenAI says it protects children across its products. The fact that state attorneys general helped write it gives it added credibility — it provides OpenAI political cover while giving the participating officials something concrete to show their constituents that progress is happening.

How Regulators Are Examining OpenAI

State attorneys general have several legal tools they can use in cases like this. They can rely on consumer protection laws, rules about charitable trusts (relevant here because OpenAI began as a nonprofit), and newer state laws protecting children online — laws modeled on a federal rule called COPPA but with stronger enforcement power. When 44 states coordinated that warning letter in August 2025, they likely used multiple legal angles at once, a common tactic to cover more ground.

OpenAI has made a point of cooperating with these inquiries. The company has a formal policy on raising concerns internally, dated January 2026, that prohibits retaliation against employees who surface problems or work with government investigators. That policy matters now because if any investigators later claimed OpenAI pressured witnesses, the company would face serious additional legal trouble.

What stands out about this situation is that it involves more than one regulator taking independent action. Forty-four states sending a joint warning, two states reviewing a major business restructuring, and a bipartisan task force producing a shared document show a regulatory response that operates on two levels at once — asking tough questions while also leaving room for the company to work out solutions and demonstrate compliance.

OpenAI has generally chosen to work with regulators rather than fight them. The company co-authored the child safety blueprint with the same state officials whose offices are investigating it, allowed Delaware to review its restructuring, and participated in the North Carolina task force. Whether this reflects genuine agreement with regulators or a strategy to shape the rules before the federal government steps in is an open question.

The investigation is still ongoing. As of June 13, 2026, no formal enforcement actions or conclusions have been announced publicly.