Anthropic expands Claude access and credits for scientific research

Anthropic expanded its programs for scientific users on August 27. According to the announcement, the company opened 10,000 seats through a new Claude team plan for scientists and broadened its AI for Science credit program beyond its earlier emphasis on biological sciences. Eligibility for the team plan begins with a verified principal investigator or equivalent at an academic or nonprofit research institution.

Access is becoming a managed research dependency

The program combines subscriptions, project credits, and Claude Science, the workbench Anthropic launched in June. Anthropic says researchers from other scientific fields can apply for credits, including compute-heavy projects, while biology and chemistry users continue to face model-access restrictions tied to dual-use safeguards. The source describes a separate program being developed with the US government for qualified life-science access to Mythos-class capabilities.

Research engineering teams should treat these programs as capacity with policy conditions rather than undifferentiated free compute. A lab needs a named plan owner, onboarding and offboarding rules, approved data classes, spending or usage alerts, and a reproducible record of model versions and tools. Credits that accelerate an experiment do not remove the need to preserve prompts, code, artifacts, and independent checks.

What applicants should establish first

Before moving a workflow, determine whether it contains controlled, patient, proprietary, or export-sensitive data and whether the available plan permits it. Confirm how collaborators are added, what happens when the one-year program ends, and which model classes are available for the discipline.

The immediate engineering task is to package one bounded workflow with auditable inputs and outputs. Use it to measure whether Claude reduces a real bottleneck while keeping the experiment reproducible outside the subscription program.