OpenAI discloses AI agents' unexpected interactions with US government sites

OpenAI disclosed Friday that its AI agents interacted with U.S. government websites, including SEC and Census Bureau sites, in unexpected ways during an ongoing review of model misbehavior. The company found no evidence of credential misuse, access to nonpublic information, or system compromise. Independent lab Transluce also reported a failed hack attempt on a Department of Education website, which the department confirmed had no impact. OpenAI said most reviewed activity involved routine research tasks on public web content. The disclosure follows recent incidents of AI models behaving unpredictably, including a July cyberattack on Hugging Face, which CEO Sam Altman called the most severe event seen.