Anthropic Launches Verification Program Granting Life Sciences Experts Access to Mythos and Opus Models

Anthropic has announced the Life Sciences Verification Program (LSVP), which grants life sciences experts access to the Mythos, Opus, and Sonnet models. The program applies more permissive safeguards for biological tasks than previously available and is currently in beta, initially open to teams and institutions.

Grant Types and Access Scope

LSVP is divided into two Grant types based on usage purpose and risk level.

• Standard Use: Targets most life sciences tasks (R&D, clinical, regulatory, etc.) and is available to entire teams. It involves relaxed classifiers for scientific tasks and requires annual renewal.
• High-risk Use: Fully removes safeguards for high-risk tasks blocked under Standard Use. It is limited to single research projects and requires renewal every six months. Currently, it is only available for Opus 5 and Sonnet 5, while Mythos access is restricted to a few institutions pending agreement with the US government.

Security and Monitoring Strategy

The program shifts from real-time blocking to Offline monitoring to reduce interruptions to legitimate work and detect distributed abuse attempts. LSVP traffic data is retained for 30 days and is not used for model training. Organizations define and manage their own safe usage scopes, and if unauthorized activity occurs, they follow administrator notification and response procedures.

Availability and Restrictions

Currently available on the 1st party console (API), Claude for Enterprise, and Team plan. Individual plans and third-party platforms are not yet supported, and BAA-enabled organizations are excluded due to the beta nature of the program. Customers holding PHI data must use separate non-BAA/non-HIPAA organizations.