Reuters reports OpenAI agents were hijacked to post on a German website in a previously undisclosed AI security incident

Reuters says OpenAI agents were involved in a previously undisclosed incident in which they were hijacked to interact with or post on a German website. The report points to emerging risks around autonomous AI systems being manipulated after deployment.

The incident highlights how AI agents can create new security exposure when they act across websites and services with limited human oversight. As agentic tools become more capable, companies will face pressure to improve monitoring, permissions, and containment for autonomous workflows.