The Anthropic Life Sciences Verification Program (LSVP) is 📝Anthropic's vetting program that gives verified life-science organizations access to 📝Claude models with relaxed biology safeguards, opened to public applications on September 17, 2026.

Anthropic's generally available models, including Claude Fable 5.1 and 📝Claude Opus 5.5, run safety classifiers that block much biology work, including legitimate drug discovery and research. The LSVP is the path around those blocks for qualifying teams: an organization passes a credentials, security, and ethics review, states what it will use the access for, and accepts monitoring against that statement. Anthropic developed the program in partnership with the US government, onboarded dozens of organizations through early access, and then opened it as a public beta. It is the life-sciences counterpart to the Cyber Verification Program that gates security work under 📝Project Glasswing.

Key Facts

Announced: September 17, 2026 (public beta applications); early access preceded it
Models, Standard Use: Claude Mythos 5.1, Claude Opus 5, and Claude Sonnet 5
Models, High-risk Use: Claude Opus 5 and Claude Sonnet 5 initially; Mythos high-risk grants remain limited pending US government coordination
Surfaces: Claude Console, Claude for Enterprise, and Team plans; not Pro or Max, third-party platforms, or BAA-enabled organizations
Apply: claude.com/form/life-sciences-verification-program
Named in the announcement: Xaira Therapeutics, Edison Scientific, and Manifold Bio

How It Works

Standard Use — covers most life-science work, including basic science, R&D, clinical development, manufacturing, supply chain, quality assurance, regulatory affairs, and investing; granted to a team and renewed annually.
High-risk Use — an add-on that removes life-science safeguards for one named research project, requires additional vetting, and renews every six months.
Verification — review of research credentials, security standards, and ethical research oversight before access is granted.
Obligations — the organization specifies intended use cases, accepts offline monitoring against them, keeps 30-day data retention for review of flagged activity, and responds to flags within agreed timeframes.

Why It Matters

The LSVP extends the access model Anthropic built for cybersecurity to biology, one of the two domains where its most capable models are most tightly restricted. Rather than lower safeguards for everyone, Anthropic verifies who is asking and monitors what they do, letting drug-discovery and research teams use Mythos-class and Opus-class models on work the public models refuse.

The program sits alongside Anthropic's broader push into science: on August 27, 2026, it launched a Claude Team plan for scientists offering 10,000 free or discounted seats to principal investigators at academic and nonprofit institutions, and expanded its AI for Science program with up to $50,000 in credits per project. Source: Anthropic — Expanding our support for scientists.

FAQ

What is the Anthropic Life Sciences Verification Program?

It is Anthropic's program for verified life-science organizations to use Claude models with biology safeguards relaxed, covering work such as drug discovery, research biology, clinical development, and manufacturing that the generally available models block.

Which Claude models does the LSVP cover?

Standard Use covers Claude Mythos 5.1, Claude Opus 5, and Claude Sonnet 5. High-risk Use initially covers Claude Opus 5 and Claude Sonnet 5, with Mythos high-risk grants still limited.

How do I apply to the Life Sciences Verification Program?

Organizations apply through Anthropic's application form at claude.com. Access runs through the Claude Console, Claude for Enterprise, or Team plans; individual Pro and Max subscriptions are not eligible.

What does an organization agree to in the LSVP?

It states its intended use cases, accepts monitoring of its traffic against them, keeps flagged activity available for 30 days, and commits to act on flags within agreed timeframes.

📝Claude Mythos — the model tier whose biology safeguards the program relaxes
📝Claude Opus — the Opus tier, including the Opus 5 model covered by the program
📝Claude Sonnet — the Sonnet tier, including the Sonnet 5 model covered by the program
📝How to Prompt Claude Opus 5.5 — handling biology refusals on the public models