The news
On September 17, 2026, Anthropic opened applications for its Life Sciences Verification Program (LSVP), which gives vetted life science teams its Mythos, Opus and Sonnet models with biology safeguards that block fewer requests, in exchange for credential checks and ongoing monitoring.
Anthropic said the program covers work that its generally available Fable models block, such as drug discovery, research biology, clinical development and manufacturing. It had already onboarded dozens of organizations through early access and said it expected to enroll hundreds within the first week. The program launched in beta for teams and institutions, with individual Pro and Max plans to follow.
Verified teams can apply for two kinds of grants. Standard Use grants cover Mythos 5.1, Opus 5 and Sonnet 5, plus future models as they launch; they span basic science and R&D through manufacturing, regulatory affairs and investment diligence, can extend to whole teams and renew yearly. High-risk Use grants remove all safeguards that block life sciences requests, apply to a single research project and must be renewed every six months.
High-risk grants are available for Opus 5 and Sonnet 5, but for Mythos they remain limited to a small set of entities; Anthropic said it is working with the U.S. government to make them more broadly available. Each applicant's research credentials, security standards and ethical research oversight are reviewed, and other safeguards, such as cyber classifiers, stay in place.
The program also changes how Anthropic polices misuse. Each organization's access is tied to the use cases in its application, described at a high level without sensitive information or intellectual property. Rather than rejecting risky requests one by one in real time, Anthropic said it monitors LSVP traffic offline for patterns outside that stated scope, and requires 30-day retention of that traffic to do so. It said the data is compartmentalized, cannot be used for model training and is not accessible to its own life sciences research teams.
Early participants named by Anthropic included Xaira Therapeutics, Edison Scientific and Manifold Bio. The launch followed an August 7 update in which Anthropic rewrote the classifier behind Fable 5's biology safeguards and said the change cut biology-related fallbacks, where the system switches to a less capable model after a biology query, by about 85% in its testing.
The numbers
- High-risk Use grant renewal
- Every 6 months
- Standard Use grant renewal
- Yearly
- Required retention of LSVP traffic
- 30 days
- Organizations Anthropic expected to enroll in week one
- Hundreds
- Drop in Fable 5 biology fallbacks after August 7 update (Anthropic testing)
- About 85%
Why CEOs should care
For pharma and biotech R&D leaders, the program removes a practical obstacle: legitimate work such as drug discovery being blocked or routed to a weaker model. Choose the grant that fits. Standard Use covers routine work across a team, while High-risk Use is scoped to one project and must be renewed every six months. Applications hinge on a clear, high-level description of intended use, so research heads should draft it with compliance before applying.
For CISOs and legal teams, the trade-off is retention. LSVP traffic must be kept for 30 days for monitoring, which may clash with zero-data-retention terms negotiated for other Claude use. Anthropic says the data is compartmentalized and excluded from training, but companies should confirm how retention applies to confidential compound data and any patient information, and what happens when usage is judged out of scope.
For boards, this is a governance commitment as much as a tool. Verification reviews security standards and ethical research oversight, and monitoring checks usage against the purposes a company declares. Boards should ask management who owns each grant, how scope is controlled as projects change, and what the fallback is if access is paused.
The bigger picture
Frontier AI developers are moving from blanket refusals to verified access tiers in fields where the same capabilities can help defenders or attackers. On September 1, Anthropic had named cybersecurity and life sciences as its trusted-access routes to Mythos 5.1, and its Cyber Verification Program already offered reduced cyber safeguards for defensive security work.
The LSVP's shift from real-time blocking to offline, pattern-based monitoring is a bet that serious misuse is easier to catch across many sessions than request by request, while letting legitimate researchers work with fewer interruptions.
What happened next
On September 22, Anthropic released Claude Opus 5.5. Because Standard Use grants are meant to extend to future models as they launch, grant holders should confirm when new models such as Opus 5.5 come under their grants.
Also watch for enrollment figures against Anthropic's target of hundreds of organizations in the first week, the promised expansion to individual Pro and Max plans, and wider High-risk access to Mythos, which Anthropic tied to its work with the U.S. government.
