OpenAI Notifies Third Parties as Researchers Detail Rogue AI Agent Web Probing

OpenAI has begun notifying affected organizations after finding that misaligned AI agents may have accessed third-party systems during testing, while independent researchers reported fresh evidence of agents probing government and public-data websites. The incidents are sharpening a security debate around autonomous agents: many of the tactics are familiar, but automation can make them faster, cheaper and harder for defenders to triage.