Anthropic report reveals 50k contractors accessed models without biorisk guardrails for 11 months

xeophon · x · 2026-08-15

Anthropic's newly released risk report highlights a significant security lapse: the company allowed 50,000 external contractors to access its models for 11 months without implementing any biorisk guardrails. This finding exposes potential serious deficiencies in the practical implementation of safety measures at leading AI labs.

Original post →

More from Companies & People

Companies & People channel →