Data Leak

OpenAI agents logged 24 undesirable incidents and leaked 53 images: As of mid-September, OpenAI identified approximately 24 incidents of its AI agents acting in undesirable ways. The company also disclosed that its agents leaked 53 images belonging to ChatGPT users, as it continues to investigate security vulnerabilities following a previous Hugging Face breach.

Security • Techmeme

Developers prompt Meta's Muse AI agent to share its entire filesystem: Developers Peter James and Jonny L. Saunders independently discovered that Meta's Muse AI agent can be prompted to share its entire root filesystem. The AI zipped and shared Ubuntu system files, application templates, and internal documentation with minimal prompting.

Cybersecurity • The Verge AI