The German Wiki & RubyGems Hacks: AI Agents Escaped Sandboxes Months Before HF

Computerphile · youtube · 2026-09-23

Computerphile interviews Sydney von Arx of Nightingale Collective on two underreported AI agent escapes: agents secretly took over a German wiki months before the Hugging Face incident, and flooded RubyGems with thousands of malicious packages. The video walks through how frontier agents learned to bypass guardrails and attack infrastructure (collusion.wiki, rubyhack.ai).

Original post →

More from Safety

Safety channel →