Multiple state attorneys general have launched a formal investigation into OpenAI, focusing on whether the company's products cause harm to children. This is the most serious legal challenge the company has faced so far.

The Wall Street Journal reported on June 12, 2026 that the investigation centers on claims that OpenAI knowingly released an unsafe product and ignored warnings from its own employees about potential dangers to users. The investigation followed a lawsuit filed by Florida on June 1, 2026 — the first state-level lawsuit of its kind against OpenAI, according to Reuters — which alleged that ChatGPT provided information that could help school shooters, offered guidance on self-harm, and created addictive patterns of use in young people.

Regulators have been building pressure on OpenAI for nearly a year. In August 2025, Arizona Attorney General Kris Mayes joined a coalition of 44 state attorneys general who sent a warning to major AI companies to stop what they called predatory practices targeting young users. That September, California's Attorney General Rob Bonta and Delaware's Attorney General Kathy Jennings sent a direct letter to OpenAI stating that harm to children would not be accepted as a normal cost of moving products to market quickly.

How Delaware Became a Leverage Point

Delaware played an especially important role. OpenAI was originally set up in Delaware as a nonprofit before transitioning to a more complex financial structure. That gave Delaware's Attorney General Kathy Jennings oversight authority over the company's restructuring. In October 2025, she completed her review and secured commitments from OpenAI to make safety changes as a condition of allowing the transition to happen. California's Attorney General also issued a statement on that same day, signaling that California was paying close attention.

The willingness of state attorneys general to tie child safety conditions to a corporate restructuring, rather than simply waiting for traditional lawsuits to work through courts, reveals a deliberate approach: use every available legal opening to extract concrete promises from companies before problems get worse.

A Parallel Task Force

Alongside the lawsuits, states have set up a separate coordinating body. In November 2025, attorneys general from North Carolina, Jeff Jackson and Derek Brown, launched a nationwide AI task force with representatives from both political parties. The task force brings in OpenAI, Microsoft, and other companies as participants, positioning it as a group that can produce both voluntary agreements and coordinate investigations across states.

OpenAI's public messaging has shifted during this period. In April 2026, the company released a Child Safety Blueprint — a document outlining what it commits to doing to protect children. The blueprint notably acknowledges the task force co-chaired by Jackson and Brown. Whether this document reflects actual changes to how ChatGPT works, or serves mainly as a way to respond to regulatory pressure, is something the current investigation may clarify.

What a Formal Investigation Actually Does

A formal multistate investigation is more powerful than letters or voluntary task forces. Attorneys general can issue demands that force companies to turn over documents, and they can combine what they learn across different states. For a company as large as OpenAI — which now has hundreds of millions of users — this kind of document discovery creates real operational burden.

The child safety allegations are also easier to address in court than broader complaints about AI. State laws protecting children online, consumer protection statutes, and product liability rules all provide established legal frameworks. Florida's specific claims — that ChatGPT gave people information to harm others, and that the system's design encouraged excessive use by minors — fit within legal theories that courts have dealt with before, even though the technology itself is new.

The practical question now facing OpenAI is whether the safety promises it made during the Delaware review, and the commitments laid out in its Child Safety Blueprint, match what it actually does with ChatGPT. Those public documents create a benchmark. If the company lives up to them, they demonstrate good intentions. If it does not, they make the gap between what OpenAI promised and what it delivers much clearer — and harder to explain to a judge.