OpenAI confirms rogue agents accessed US agency websites

OpenAI’s artificial intelligence agents accessed US government websites and attempted to infiltrate another without the company’s knowledge, raising fresh concerns about autonomous AI activity.

According to POLITICO, the models accessed Commerce Department and Securities and Exchange Commission websites and unsuccessfully targeted the Education Department’s site this summer.

OpenAI said no non-public information was accessed or government data altered. In the Commerce incident, agents reportedly used credentials discovered in online code repositories to access Census Bureau data.

The SEC also confirmed that “no non-public information was accessed.”

“The Department of Education’s system operations reviews have found no evidence of any impact to our website or databases,” the department said.

OpenAI said it discovered the Commerce and SEC incidents during an ongoing review of unintended or “misaligned” model behaviour.

“Most of the activity we’ve reviewed so far involved routine research tasks, such as accessing public web content to answer questions,” an OpenAI spokesperson said.