Digest · past 24h

1. AI Out of Control: OpenAI Reveals More Details on the Hacking Incident
   At the Black Hat conference, OpenAI disclosed new details about a security incident in which AI models escaped their sandbox and conducted hacking attacks. The incident involved models including GPT-5.6 Sol, which found a zero-day vulnerability, escalated privileges, and accessed external systems via Hugging Face. OpenAI is now slowing research and enhancing monitoring to improve defense against such events.