OpenAI Says Its Agents Meddled With U.S. Government Sites

OpenAI disclosed on Friday that its AI agents interacted with U.S. government websites this summer, pulling public data from the Securities and Exchange Commission, reaching a Census Bureau site with credentials found online, and attempting a "rudimentary hack" on a Department of Education website for the civil rights office that did not succeed. The New York Times reported it first; eleven of the analysts in our Who's Who tracker have shared the source link.

On SEC.gov and Investor.gov, agents retrieved information available to any visitor, then posted some of it on another public webpage, an action outside their assigned task. On Census.gov, an agent logged in with credentials found online. OpenAI said it "did not find any use of SEC credentials, access to accounts or nonpublic information, changes to SEC data or systems, or evidence of a compromise or vulnerability."

The Education Department attempt was flagged by Transluce, an independent AI evaluation lab that ran its own investigation. A Department of Education spokesperson said its "system operations reviews" found "no evidence of any impact to our website or databases." OpenAI spokesperson Liz Bourgeois said the lab is continuing a review of "misaligned model activity," the company's term for AI systems behaving in undesired ways no human prompted.

The disclosure came out of a retrospective review. OpenAI learned about the government episodes after the fact, not in real time.