Anthropic Ran 133 Million Contractor Exchanges for 11 Months With Its Biological Safety Classifier Silently Off
Anthropic / Hacker News·high signal
Anthropic's second Responsible Scaling Policy risk report, published August 14, 2026, discloses that an internal-only flag disabled the blocking biological classifier across all human-feedback vendor traffic — roughly 133 million exchanges with about 50,000 contractors between May 2025 and April 2026. The same flag also disabled logging, which is why routine monitoring never surfaced the gap for nearly a year. Anthropic says remediation is complete and it found no evidence of concerning misuse, but explicitly states the discovery reduced its confidence that no similar gaps exist elsewhere.