Announced on September 17, 2026, the Life Sciences Verification Program (LSVP) allows verified institutional researchers to apply for more permissive biology-related classifiers and safeguards than those built into generally available Fable models. The program covers Anthropic's Mythos, Opus and Sonnet models for workflows described by the company as drug discovery, research biology, clinical development and manufacturing. Anthropic said it had already onboarded dozens of organizations through early access and was opening applications to the wider life-science community.

Verification requires applicants to undergo review of their research credentials, organizational security standards and ethical research oversight. Access is split into two primary operational tiers rather than granting open-ended model permissions.

Two-Tiered Access Model

The program distinguishes between institutional baseline permissions and project-specific grants:

  • Standard Use: Designed for the majority of biological research tasks, Standard Use grants apply team-wide, renew annually and cover Mythos 5.1, Opus 5 and Sonnet 5, with Anthropic indicating they will include future models at launch.
  • High-risk Use: Offered as an add-on grant for activities blocked under Standard Use, this tier is scoped strictly to an individual research project and requires renewal every six months. It removes the safeguards that block life-science prompts covered under the grant.

Model availability differs within the High-risk tier. Anthropic says High-risk grants for Opus 5 and Sonnet 5 are available upon approval, but High-risk access for Mythos remains limited to a small set of entities subject to extra vetting while the company collaborates with the U.S. government. Non-biological guardrails, including cybersecurity classifiers, remain active across all LSVP accounts.

Data Isolation and Surface Restrictions

Anthropic states that LSVP traffic is retained for 30 days for offline safety monitoring. The company says the retained data is compartmentalized, cannot be used for model training and cannot be viewed by members of its internal life-sciences research teams.

Operating behavior also depends on the product interface:

  • In the direct API and Claude Science, users can switch between active grants natively.
  • In Claude.ai and Claude Code, accounts are constrained to a preselected default grant, with an exception for Claude Code when authenticated via the direct API.

Launch Boundaries and Exclusions

The beta rollout contains strict access boundaries. LSVP is limited to Anthropic's first-party API console alongside Claude Enterprise and Team subscription tiers. It is not available to individual account holders, users accessing Anthropic models through third-party cloud platforms or organizations operating under Business Associate Agreements (BAAs) during the initial beta phase.