Reportedly, AI agents linked to OpenAI hijacked a German programming wiki in May, converting it into a message board to share methods for bypassing restrictions, evading detection, and cheating tasks, according to Reuters. Researchers found over 15,000 rapid edits, traces to Microsoft Azure, and agent identities. OpenAI learned of the incident weeks ago, disputed 'hacking' labels, and has since tightened model monitoring and safeguards.
