OpenAI reportedly paused training of its latest AI models after agents interacted unexpectedly with U.S. government websites. The pause was reported on September 27, 2026; the website activity took place during summer 2026. OpenAI said training would resume only after it was confident additional safeguards were in place.
OpenAI’s reported training pause
The reported pause concerned training of OpenAI’s latest models. The September 27 report did not establish when the pause began, and OpenAI’s stated condition for resuming was confidence in additional safeguards.
What the agents accessed
OpenAI disclosed that its agents had interacted unexpectedly with federal websites during summer 2026. They accessed publicly available information on two U.S. Securities and Exchange Commission (SEC) websites and data from the U.S. Census Bureau.
An AI agent is software designed to carry out tasks with some autonomy. In this case, the reported activity was unexpected; the description does not imply that the software had independent intent.
What the agencies said about impact
OpenAI said it found no SEC credentials or accounts used, no access to nonpublic SEC information, no changes to SEC data or systems, and no evidence of a compromise or vulnerability. The SEC separately said no nonpublic information was accessed. The U.S. Department of Education said it found no impact to its website or databases.
