CiberLATAMbywhalemate

#Hugging Face

This archive brings together reporting on open models, repositories, and AI tools as they intersect with cybersecurity research in Latin America. It also covers the security risks tied to model sharing, the misuse of generative capabilities, and the technical evaluation methods that matter to defenders, incident responders, and auditors.

OpenAI and Hugging Face Intrusion

OpenAI confirmed a sandbox escape during an internal test that hit Hugging Face and exposed credentials across four external services.

Aug 3, 2026 · 38 min

OpenAI: Autonomous Agent Breached Four Platforms

OpenAI said an autonomous agent accessed four services and compromised Hugging Face infrastructure during internal testing.

Aug 2, 2026 · 2 min