Anthropic loosens Claude's biosafety limits for verified researchers
The Life Sciences Verification Program trades Claude's default biology-related blocks for credential checks and after-the-fact monitoring, developed in coordination with the US government, with a higher-risk tier that removes those blocks entirely.
Anthropic launched the Life Sciences Verification Program (LSVP) on September 17. The invite-only track gives vetted biology researchers and institutions access to Claude models, including Claude Mythos, Opus and Sonnet, with default biosafety guardrails relaxed for their work. Applicants are reviewed on "research credentials, security standards, and ethical research oversight." The program initially targets teams and institutions, and Anthropic says it plans to extend it to individual Pro and Max plans later.
Two tiers
Standard Use removes biology-specific blocks and keeps other safeguards, such as cyber-related classifiers, in place. High-risk Use "removes all safeguards that block life sciences requests" for specific projects that clear additional vetting. In both tiers, Anthropic moves misuse monitoring from real-time blocking to offline pattern analysis, with 30 days of data retention for anything flagged. Anthropic says it is working with the US government to expand High-risk access to Claude Mythos beyond the currently limited set of entities.
A verified gate in place of a blanket block
Biology research has always been the hardest case for AI safety filters. The knowledge that helps a vaccine researcher can help someone with worse intentions, and a hard block frustrates the first to stop the second. Anthropic is betting a verified-identity gate handles that better. "Dozens of organizations" joined during early access, and Anthropic expects to enroll "hundreds of organizations" in the first week of the broader rollout. If your life-sciences organization already pays for Claude, apply and check the change against your internal usage: it changes what the model will do for a verified account.