Anthropic praised the White House’s “Winning the Race: America’s AI Action Plan,” published on July 23, 2025, for prioritizing AI infrastructure, federal adoption, safety testing and security coordination. The company said many recommendations align with its earlier submission to the Office of Science and Technology Policy (OSTP), but argued that binding transparency requirements for AI development and sustained export controls are still essential to secure US leadership.
Accelerating infrastructure and federal adoption
Anthropic welcomed the plan’s emphasis on accelerating AI infrastructure and streamlining data center and energy permitting, echoing its March submission to OSTP. The company warned that insufficient domestic power capacity could push developers to relocate overseas and expose sensitive technology to foreign adversaries. Anthropic cited its own “Build AI in America” report as a blueprint of steps the administration can take to expand domestic energy capacity and said it looks forward to cooperating on implementation.
The Action Plan’s proposals to increase federal AI adoption also reflect Anthropic’s priorities, including:
- directing the Office of Management and Budget (OMB) to address resource constraints, procurement limitations, and programmatic obstacles to federal AI deployment;
- launching a Request for Information (RFI) to identify federal regulations that hinder AI innovation, with OMB coordinating reforms;
- updating federal procurement standards to remove barriers preventing agencies from deploying AI systems;
- promoting public–private collaboration to expand AI use in defense and national security applications.
Broadening access to AI’s benefits
Anthropic supports the plan’s focus on broad participation in and benefit from AI. It welcomed continuation of the National AI Research Resource (NAIRR) pilot, which aims to give students and researchers nationwide access to AI research resources; Anthropic said it has long supported NAIRR and is a partner in the pilot. The Action Plan’s emphasis on rapid retraining for displaced workers and pre-apprenticeship AI programs was also praised as a corrective to prior technological transitions.
To complement these measures, Anthropic pointed to its Economic Index and Economic Futures Program as tools to help researchers and policymakers understand how AI is reshaping the economy and to ensure benefits are broadly shared and risks are managed.
Supporting secure AI development
Anthropic agreed with the plan’s focus on defending against misuse of powerful AI models and preparing for future AI risks. It specifically commended the administration’s prioritization of research into AI interpretability, control systems, and adversarial robustness. These research areas, the company said, are necessary to manage increasingly capable systems.
The company also welcomed the Action Plan’s affirmation of the National Institute of Standards and Technology’s Center for AI Standards and Innovation (CAISI) and said it expects to continue close collaboration. Anthropic warned that advanced AI systems are showing capability improvements that are relevant to biological weapons development; it said CAISI has played a leading role in developing testing and evaluation capabilities to address such risks and urged focusing resources on the most acute national-security threats.
The need for a national transparency standard
Beyond testing, Anthropic argued for basic AI development transparency requirements, such as public reporting on safety testing and capability assessments. The company said leading model developers should be held to publicly verifiable standards for assessing and managing catastrophic risks posed by their systems, and pointed to its proposed framework for frontier model transparency. Anthropic expressed disappointment that the Action Plan did not do more on this subject.
Anthropic noted that leading labs — including Anthropic, OpenAI, and Google DeepMind — have implemented voluntary safety frameworks, showing responsible development can coexist with innovation. As an example, Anthropic said it proactively activated ASL-3 protections upon launching Claude Opus 4 to prevent misuse related to chemical, biological, radiological, and nuclear (CBRN) weapons development. The company argued such precautions do not hamper innovation but help build more reliable systems.
Anthropic shared the administration’s concern that overly prescriptive regulation could produce an inconsistent and burdensome patchwork of laws. Ideally, it said, transparency requirements would be set by a single national standard. However, Anthropic reiterated its opposition to proposals for a ten-year moratorium on state AI laws, calling such a moratorium too blunt an instrument and arguing states should still be able to act if the federal government does not.
Maintaining strong export controls
The Action Plan states that denying foreign adversaries access to advanced AI compute is both a geostrategic competition and national security issue. Anthropic strongly agreed and expressed concern about the administration’s recent reversal on export controls for Nvidia H20 chips to China.
Anthropic emphasized that AI capability scales with compute, energy, and data. It noted that inference capability in modern reasoning models is limited in part by a chip’s memory bandwidth. While the H20’s raw compute may be exceeded by some Huawei chips, Anthropic cited testimony by Commerce Secretary Lutnick and Under Secretary Kessler that Huawei continues to struggle with production volume and no domestically produced Chinese chip matches the H20’s memory bandwidth. As a result, the company argued, H20 exports to China would provide computing capabilities otherwise unavailable to Chinese firms and could compensate for China’s chip shortage.
For that reason, Anthropic urged the administration to maintain controls on the H20. The company said such controls align with the export-control recommendations in the Action Plan and are essential to preserving and extending American AI advantage.
Looking ahead
Anthropic concluded that the overlap between its recommendations and the Action Plan shows a shared understanding of AI’s transformative potential and the urgent steps needed to sustain US leadership. The firm said it looks forward to working with the administration to implement the proposals while ensuring attention to catastrophic risks and strong export controls.
For further detail Anthropic referred to its full OSTP submission, ongoing work on responsible AI development, and its recent report on increasing domestic energy capacity.
