Anthropic commits $150 million in Claude access to US science agencies

Anthropic will provide $150 million in AI tools, credits, training, and technical support to the Genesis Mission over three years. The federal program plans to use AI to accelerate scientific and technological research across more than 15 agencies, including NASA, the National Institutes of Health, and the National Science Foundation.

The company announced its $150 million commitment at the White House Office of Science and Technology Policy’s Science: A New Golden Age Summit. The Department of Energy announced 12 Phase II Genesis Mission project awards worth $159 million at the same event.

The two announcements represent different forms of support and do not constitute a single funding pool. The Energy Department is financing projects, while Anthropic is valuing its contribution through products and services.

Inside Anthropic’s package

Anthropic describes its contribution as in-kind support rather than a cash grant to the government. The package includes:

• Model access: Claude models for participating researchers and agencies.
• Developer tools: Claude Code, Anthropic’s coding tool, for programming and software-development work.
• API credits: Funding for several hundred projects to integrate Claude into applications and research workflows.
• Technical support: Training, onboarding, and engineering assistance for agencies and national laboratories.
• Priority research: Collaboration in fields that include fusion energy and quantum computing.

API access allows research teams to call Claude programmatically from internal software, data pipelines, and automated experiments. Claude Code adds support for tasks such as writing, reviewing, and debugging research software.

Anthropic has not disclosed the available model tiers, token allocations, rate limits, data-retention terms, deployment environments, or rules for handling controlled and classified information. Those details will determine which projects can advance beyond limited pilots and how agencies can integrate Claude with sensitive systems.

Genesis expands beyond the national labs

The Trump administration created the Genesis Mission by executive order in November 2025. The Department of Energy leads the initiative, with Under Secretary for Science Dario Gil directing it. Its remit spans energy, space, health, and other federally supported research fields.

Anthropic’s work with federal laboratories began before the mission’s formal launch. Lawrence Livermore National Laboratory expanded Claude access to about 10,000 scientists, researchers, and staff in July 2025. Anthropic announced a broader Energy Department partnership in December 2025 covering energy, biological and life sciences, and scientific productivity.

The new commitment extends that work beyond the Energy Department and national laboratories. Participating agencies could use a common set of models and developer tools across fields ranging from biomedical research to space science.

Anthropic targets scientific workflows

Anthropic has spent the past year assembling products and programs for researchers. It launched Claude Science, a research workbench connected to commonly used scientific libraries and tools, and offered 10,000 free or discounted Claude seats to academic scientists.

The company also opened a research preview of the Model Hardware Standard, a proposed specification for connecting AI agents to laboratory instruments. A shared interface could let software operate compatible equipment without custom integration for every device, provided laboratories establish appropriate safety controls and approval steps.

Federal adoption gives Anthropic a large base of demanding users and a reference for other research and regulated markets. It can also make Claude costly to replace once agencies build software, documentation, and training around its APIs. Anthropic has not said the Genesis arrangement is exclusive, leaving agencies free to evaluate competing models under their procurement and security requirements.

Results will determine the value

Claude can help researchers search literature, write code, analyze text, and generate hypotheses. Scientific claims still require validation against data, experiments, and domain expertise because model outputs can contain fabricated citations, faulty reasoning, or plausible but incorrect conclusions.

Anthropic’s announcement does not identify the participating projects or define outcome measures. Agencies will need evaluation plans that track:

• Accuracy against expert-reviewed baselines
• Reproducibility and provenance of model-assisted results
• Time and computing costs compared with existing workflows
• Security incidents, data exposure, and unauthorized tool use
• Portability across models and vendors
• Research outcomes that survive peer review or experimental validation

Researchers and journals will also need policies for disclosing model use, citing generated material, preserving prompts and outputs, and protecting sensitive data. The three-year program gives Anthropic broad access to federal scientific workflows, while measurable research results will show whether that access produces gains beyond faster drafting and coding.