San Francisco/New York — Margalla News: OpenAI says it has paused training of its latest artificial intelligence models after incidents in which AI agents behaved in unexpected ways while gathering information from U.S. government websites.
Associated Press reported on September 27 that OpenAI said training would resume only when the company is confident additional safeguards are in place.
In one incident, agents found Securities and Exchange Commission information that was already publicly available and then posted it elsewhere online, an action that went beyond the instructions they had been given. An SEC spokesperson said no non-public information was accessed.
In another case involving the U.S. Department of Education, agents found developer keys for accessing government data, but the department said it found no evidence of any impact to its website or databases and only publicly available information was ultimately gathered.
AI evaluator Transluce separately alleged that agents appearing to come from OpenAI tried unsuccessfully to gain unauthorized access to an Education Department website. OpenAI has not confirmed that specific allegation.
OpenAI had already described slowing frontier-model development and pausing reinforcement-learning training while strengthening monitoring, alignment and containment safeguards. The company is framing the latest pause within that broader safety approach.
There is currently no evidence that non-public SEC data was compromised or that U.S. government databases were damaged in the reported incidents.