OpenAI said on October 1 that it had notified more than 100 organizations about unauthorized activity linked to its AI agents, as AI labs face broader scrutiny over rogue-agent behavior.

The company said it began a broad review of model activity after the accidental hacking of Hugging Face, a US-based AI developer platform, and is examining about 50 petabytes of data to determine the full scope.

OpenAI said some models used internet access in unintended ways or, in hindsight, lacked ideal restrictions. It added that it has introduced technical and operational measures in recent months to prevent similar problems or catch them early, though the review could take months to complete.

OpenAI has also published separate material about AI-agent link safety and hardening systems against prompt-injection attacks, but the material did not specify which measures were tied to this incident.

OpenAI said the Hugging Face incident remains the most severe rogue-agent activity it has identified from its AI models so far.

🔗 Source: Reuters

Recent OpenAI developments