Anthropic opens institutional biology access with mandatory data retention

Organizations undergo review of credentials, security and research oversight. Launch access covered Anthropic’s API and Claude Team and Enterprise, excluding individual Pro/Max subscriptions and third-party platforms. [1] The institution, rather than an individual account holder, is the unit of admission.

The program distinguishes broad team access from additional approval for particular projects. Standard grants renew annually; the more permissive project-specific category renews every six months. These are access permissions, not research funding. Other safeguards remain in place. [1] The application page, rechecked on September 28, asks an organization’s administrator to register interest and explicitly says that applying does not guarantee admission. [2]

Data handling is a consequential condition. Program traffic requires thirty-day retention; Anthropic says it is compartmentalized and unavailable for training or its life-sciences researchers. Organizations configured under a business associate agreement for protected health information were excluded from the beta. [1] Those terms deserve review before a team decides which work belongs in the program.

A useful application would start with the actual workflow: what information the model receives, which legitimate task is being blocked and who is responsible for reviewing the output. That makes the access request assessable and exposes any mismatch with the institution’s obligations. Approval could reduce interruptions in eligible research, but it would not establish the scientific validity of an answer. The opportunity is a more explicit relationship between a research organization and its AI provider, with permission, monitoring and evidence quality treated as separate responsibilities. A credible request makes that division of responsibility concrete.