Verified access in two tiers
Anthropic's Life Sciences Verification Program, launched in beta for teams and institutions on September 17, 2026, gives verified life-science professionals access to its Mythos, Opus and Sonnet models with safeguards more permissive for biology work than those on its generally available Fable models. Applicants are reviewed for research credentials, security standards and ethical oversight; Anthropic said dozens of organizations had already joined through early access. [1]
Standard Use grants cover whole teams, renew yearly and apply to Mythos 5.1, Opus 5 and Sonnet 5, with classifiers tuned to allow most biology research and development. High-risk Use grants are project-specific add-ons, renewed every six months, that remove all safeguards blocking life-science requests. They were available for Opus 5 and Sonnet 5 at launch; for Mythos, Anthropic said it was working with the U.S. government and would limit them to a small set of entities with additional vetting. Cyber classifiers and other non-biology safeguards stay in place. [1]
Monitoring against a stated purpose
Anthropic said it designed the program's safeguards against compromised accounts, insider threats and agents taking dangerous actions, by monitoring each organization's usage against the use it declared, and it cited attempts in its recent threat report that could support biological weapons development. [1]
Atlas interpretation: The program replaces a blanket refusal with vetting and monitoring, which is how the most capable biology help reaches researchers without being on by default. It also moves part of the judgment about safe use to the approved organizations themselves. The next day, Anthropic's head of life sciences confirmed to Reuters its own biology wet lab. [1][2]
Sources
- Introducing the Life Sciences Verification Program
Anthropic · Sep 17, 2026
- Anthropic quietly sets up biology lab as it ramps AI drug program: Reuters
CNBC · Sep 18, 2026