Anthropic Opens Life Sciences Verification Program for Biology Researchers Using Claude
A two-tier verified access system calibrates Claude's safeguards for drug discovery, vaccine research, and clinical development
Anthropic Life Sciences Verification Program is now open in beta, giving biology researchers at academic labs, pharmaceutical companies, and life science startups a path to use Claude's Mythos, Opus, and Sonnet models with safeguards calibrated for scientific work. No general availability yet. Teams and enterprise plans can apply now.
The Dual-Use Problem Driving the Program
Biology is one of the few research domains where legitimate work and potential harm look nearly identical from the outside. Developing a vaccine and exploiting a pathogen can draw on the same underlying knowledge - the intent and institutional context are what separate them. Standard AI safeguards, built for general consumer use, cannot make that distinction without verified context about who is asking and why.
Anthropic addresses this directly in the program announcement: "valid research (e.g. developing vaccines) and harmful pursuits can be difficult to distinguish." Rather than restricting biology broadly, the program routes researchers through credential verification before adjusting what the model will engage with.
Two Tiers, Two Risk Levels
Standard Use covers the bulk of life science work - basic research, R&D, supply chain operations, clinical development, and regulatory affairs. Access renews annually and covers entire teams.
High-risk Use is for research areas that require closer vetting. Access is project-specific rather than organization-wide, renewals run every six months, and approval removes all life sciences safeguards from the model entirely. Anthropic has not published a list of what qualifies for that tier - each application goes through individual review.
What Verification Requires
Joining the Anthropic Life Sciences Verification Program requires submitting research credentials, documented security standards, and evidence of ethical oversight before Anthropic grants access. Once approved, usage is monitored offline against the stated use cases, with a 30-day data retention window. Anthropic reviews each application individually.
Where the Program Runs
LSVP access works across four products: Claude Science, Anthropic's research workbench for scientists, Claude.ai, Claude Code, and the API. All three models - Mythos, Opus, and Sonnet - are included, giving verified researchers access to the full current model lineup rather than a restricted tier.
Anthropic previously created a verified-access channel for US government organizations using Mythos 5, carving out controlled access for sensitive work rather than limiting the model outright. LSVP applies that same logic to scientific research. Verification as the mechanism for expanding capability is a different posture from restriction-as-default, and it matters for fields where over-restriction costs real research progress.
Anthropic plans to expand the Anthropic Life Sciences Verification Program to individual accounts but has not announced a timeline. Applications are open now for eligible teams and enterprise organizations.