Anthropic launches life sciences verification program to enhance AI research access

Anthropic has launched the Life Sciences Verification Program (LSVP), offering life science professionals access to its Mythos, Opus, and Sonnet models with enhanced safeguards tailored for biology-related tasks. The program, initially in beta, is open to teams and institutions, with plans to expand to individual Pro and Max users over time. LSVP aims to address limitations in its generally-available Fable models by enabling tasks such as drug discovery and clinical development. The program includes two grant types: Standard Use for most life science work and High-risk Use for projects requiring fewer restrictions. High-risk grants, available for Opus 5 and Sonnet 5, are limited to specific entities due to ongoing government collaboration. All LSVP traffic is monitored for 30 days, with data retention required for offline analysis. Anthropic emphasized that this data is not used for model training or accessed by its life sciences research teams. The initiative follows recent security incidents involving unauthorized access to computer systems, prompting an independent review. Anthropic also announced a research preview of the Model Hardware Standard (MHS) for select labs and manufacturers. The LSVP is available via API and Claude Enterprise, with support for individual plans and third-party platforms expected in the future.