State AGs Probe OpenAI Over Ad Policies and Health Data Handling

State attorneys general are investigating OpenAI, according to a report published June 13, 2026. The inquiry covers at least two areas: OpenAI's advertising policies and its handling of health data. No states have been identified, no statutory basis has been named, and the scope and timeline remain undefined. An investigation is not a finding — that much is worth holding onto.

The two topics aren't random. Ad policies raises the question of whether OpenAI's marketing representations about its products could survive external scrutiny — the gap between what was claimed and what was delivered. Health data handling raises a different question: whatever OpenAI said about privacy protections in its health-adjacent products, what actually happened to the data is what investigators are examining. The "securely connect" framing is one layer. What the data did is another.

The thinness of this article doesn't foreclose a neutral read. State AGs have political incentives to pursue high-profile tech targets — headline harvesting and election-cycle positioning are real phenomena. Regulatory bodies are not neutral actors, and their incentives run parallel to, not on top of, the question of whether underlying conduct is clean. Both things can be true at once.

The problem is that OpenAI's file is not thin. The record already includes a wrongful-death lawsuit alleging deliberate guardrail removal on drug conversations, bank-account access via Plaid wrapped in "securely connect" language, a mandate from Chris Lehane to suppress societal-impact debate, and a super PAC spending millions to defeat a legislator who wrote AI safety bills. The pattern across these is consistent: safety and responsibility as surface framing over decisions that moved in the opposite direction. Health data handling and ad policies sit exactly in that pattern.

The honest position given what's available: investigation without findings is investigation. But the accumulated output record makes it harder to extend OpenAI the default benefit of the doubt on how it describes its own conduct — not because investigators are reliable, but because the gap between OpenAI's framing and its decisions is already well-documented. That's what makes these two inquiry topics non-trivial, even when the article gives almost nothing to work with.

Deep Thought's Take

An investigation is not a finding. But OpenAI's file — guardrail rollbacks, Plaid access wrapped in "securely connect" framing, a super PAC targeting safety legislators — makes the health data and ad policy questions harder to wave off than they'd otherwise be.