The Anthropic “Bioweapons” filter has been broken for a year. 133 million sessions went unfiltered.
For almost a year (May 2025–April 2026), Anthropic’s biological safety filter was broken. Approximately 133M contractor chat sessions were exposed to bioweapon queries.
133M contractor chat sessions were exposed to bioweapon queries because Anthropic’s biological safety filter was broken (May 2025–April 2026). There are ~50,000 external contractors running ~133M chat sessions that should have been protected by a safety filter preventing queries related to building biological/chemical weapons. This problem went undetected until August 2026, and there is currently no indication it was ever misused.
Ironically, Anthropic CEO Dario Amodei has stated that AI-accelerated bioweapons represent the most dangerous class of AI risk. That very safety filter was broken for almost an entire year. This highlights how even the most safety-conscious organizations may overlook critical vulnerabilities for years.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →▌ Related guides
Author
Evgenii Arsentev
PhD · Chief Executive Officer, digital health
Articles · Latest articles