The Pulse
Anthropic Opens Biology Access to Vetted Research Teams
Anthropic is opening applications for its Life Sciences Verification Program, which gives vetted teams access to Claude models with fewer biology-related restrictions. The beta covers drug discovery, research biology, clinical development a
AI.info Team ·
Anthropic is opening its biology capabilities to a wider group of researchers while moving some safety enforcement away from the model’s immediate responses and into organization-level monitoring.
The company announced the Life Sciences Verification Program on September 17, opening applications to the broader life sciences community after onboarding dozens of organizations through an early-access program. The beta initially targets teams and institutions, with access to Anthropic’s Mythos, Opus and Sonnet models under safeguards that are more permissive for biology-related work than those on its generally available Fable models.
The move addresses a direct conflict in Anthropic’s product strategy: scientists need models that can discuss real research and development, while the same capabilities could assist harmful biological work. Anthropic’s answer is to limit access to vetted organizations, tie permissions to declared projects and monitor activity for signs of misuse.
Anthropic Draws a New Line Around Biology Access
The program covers work that Fable models currently block or redirect, including drug discovery, research biology, clinical development and manufacturing. Anthropic says the program is intended for teams ranging from academic laboratories to startups and pharmaceutical companies.
Applicants must undergo a review of their research credentials, security standards and ethical oversight. Approved organizations can then request one of two access levels: Standard Use or High-risk Use.
Standard Use grants are designed for most life sciences research and development workflows. They can cover an entire team, renew annually and apply across areas including basic science, supply chains, manufacturing, clinical development, quality assurance, regulatory affairs and investment analysis.
High-Risk Projects Face Separate Review
High-risk Use grants apply to work that Anthropic considers more likely to be misused. The grants remove the life sciences safeguards that block those requests, but they apply to a single research project rather than a full team and must be renewed every six months.
Anthropic gives as an example a project studying how a specific family of viral vectors is recognized by human immune pathways. High-risk grants for Claude Opus 5 and Claude Sonnet 5 are available at launch. The company says high-risk access for Claude Mythos remains limited to a small group of entities receiving additional review, in cooperation with the U.S. government.
Other protections remain active under the program. Anthropic says its cybersecurity classifiers continue to apply, even when an organization receives a biology grant.
Monitoring Replaces Some Real-Time Blocking
The program’s main technical change is a shift from blocking individual requests in real time to examining patterns of activity after the fact. Anthropic says that approach should reduce interruptions for legitimate research while giving its systems more context when assessing possible misuse.
That tradeoff comes with a data-retention requirement. Anthropic will retain data associated with Life Sciences Verification Program traffic for 30 days to support monitoring and review. The company says the data is compartmentalized, cannot be used to train models and cannot be accessed by its life sciences research teams.
Anthropic says the safeguards focus on three threats: account compromise, insider misuse and dangerous actions by AI agents operating across long tasks or in groups. Access is tied to the uses described in an organization’s application, and the company will monitor traffic for activity outside that approved scope.
If the system identifies unauthorized activity, Anthropic says it can alert organization administrators, who must act within agreed timeframes to investigate and address the incident. The company is also examining how the program could work with its Enterprise Frontier Safeguards, which are designed to let customers store data in their own cloud infrastructure.
Drug Discovery Companies Join the Early Program
Anthropic’s early participants include Xaira Therapeutics, Edison Scientific and Manifold Bio. Xaira said it plans to use Anthropic’s models across its drug discovery work.
“We’re excited to put Anthropic's frontier models to work across our drug discovery engine, and we believe pairing trusted access with intelligence is the right way to realize AI’s promise in biology.”
Xaira Therapeutics
Edison Scientific said the program would let it use Anthropic’s most capable models for work focused on science and medicine. Manifold Bio also welcomed the company’s decision to pair broader access with organizational accountability.
Applications Open, With Limits Still in Place
Organizations can apply through Anthropic’s program. The company expects to enroll hundreds of organizations during the first week and expand the program further in the following weeks.
At launch, access is available through Anthropic’s own API console, Claude for Enterprise and Claude Team plans. The program does not yet support individual plans or third-party platforms. It is also unavailable to organizations using business associate agreements, meaning customers handling protected health information must use separate non-BAA organizations.
Anthropic has not removed all restrictions from its broader Claude product line. Fable models still route many professional biology and drug-development requests to less capable Opus models, while Mythos access remains restricted to vetted organizations. The Life Sciences Verification Program is the company’s mechanism for widening access without making those capabilities generally available.