Anthropic Bio-Weapons Filter Down for a Year, Exposing 133 Million Requests
The Decoder · rss · 2026-08-16
Anthropic revealed in a safety report that its internal filtering system for biological and chemical weapons risks was inactive for nearly a year.
Incident Details:
- Duration: The filtering system was down for almost a year.
- Scale of Exposure: Approximately 50,000 external feedback contractors conducted around 133 million unfiltered interactions with the models during this period.
- Risk Implications: Users potentially accessed dangerous information regarding bio-weapon creation without safety guardrails.
This incident highlights the challenges in maintaining the stability of AI safety infrastructures and the oversight required for red-teaming and contractor workflows.
More from Safety
- METR outlines three blind spots in public knowledge of AI lab internal agent risks — CFGeek · 2026-10-16
- Open-source rizin-mcp lets LLMs run local binary and malware analysis via Rizin — Pure_Emergency1424 · 2026-10-16
- METR researcher: scariest scenario is agents actively subverting their own safety systems — deliprao · 2026-10-16
- Cryptographer Matthew Green: Pull Critical Infrastructure Firmware Now — matthew_d_green · 2026-10-16
- Kernel-Level Agent Sandboxing With macOS Seatbelt, No Docker Needed — mostly_deterministic · 2026-10-16
- Stop giving agents raw API keys: prepaid balances and per-credential caps — Cold_Hall_5384 · 2026-10-16