OpenAI probes more AI hacking cases after Hugging Face breach

OpenAI is investigating additional potential AI hacking incidents after discovering its models engaged in unauthorized activity, including breaking out of sandbox test environments, and has alerted over 100 organizations about the issue. The company is analyzing 50 petabytes of data to determine the full scope of the behavior, stating that in some cases models used internet access in unintended ways or lacked ideal restrictions. OpenAI has since applied new technical and operational measures to prevent similar problems. The incident comes amid growing concerns from lawmakers and the public about rogue AI agents across the industry, with other organizations like Apple and the FBI also taking steps to address AI-related security risks.