Anthropic's Statement on the U.S. AI Action Plan

America’s AI Action Plan, released by the White House on July 23, 2025, is a comprehensive federal strategy featuring more than 90 specific policy actions designed to secure U.S. global leadership in artificial intelligence (AI). The plan is built on three main pillars: accelerating AI innovation, building American AI infrastructure, and leading in international AI diplomacy and security. It focuses on removing regulatory barriers, supporting open AI models, expanding access to AI resources for the private sector and academia, accelerating the construction of data centers and semiconductor factories, creating jobs, and protecting against risks such as the misuse of AI. Key principles include empowering American workers, ensuring objectivity and freedom of speech in AI models, and defending against risks, including by excluding diversity, inclusion, and climate change requirements from federal AI standards.

Accelerating AI Infrastructure

The White House recently released a plan titled "Winning the Race: America’s AI Action Plan," which focuses on maintaining America’s advantage in AI development. Anthropic welcomes this plan because it emphasizes accelerating the construction of AI infrastructure and federal adoption, as well as strengthening safety testing and the coordination of safety measures. Many recommendations in the plan reflect Anthropic’s response to an earlier March 2025 request for information from the Office of Science and Technology Policy (OSTP).

Anthropic appreciates the administration’s commitment to streamlining permitting for data centers and energy infrastructure to address AI’s energy needs. As Anthropic stated in its response to OSTP and at the Pennsylvania Energy and Innovation Summit, without sufficient domestic energy capacity, American AI developers could be forced to move operations abroad, potentially exposing sensitive technologies to foreign adversaries. Anthropic’s recently released report, "Build AI in America," details steps the administration can take to accelerate the construction of national AI infrastructure, and Anthropic looks forward to working with the administration to expand domestic energy capacity.

The plan also proposes increasing AI adoption across the federal government, including tasking the Office of Management and Budget (OMB) with addressing resource constraints, procurement restrictions, and programmatic barriers. It also launches a request for information (RFI) to identify federal regulations that hinder AI innovation, with OMB coordinating reform efforts. Updates to federal procurement standards are intended to remove barriers that prevent agencies from deploying AI systems and to support AI adoption in defense and national security through public-private collaboration.

Democratizing the Benefits of AI

Anthropic agrees with the plan’s emphasis on ensuring broad participation in and benefits from the continued development and deployment of AI. Continuing the National AI Research Resource (NAIRR) pilot program ensures that students and researchers across the country can contribute to progress at the frontiers of AI. Anthropic has long supported NAIRR and is proud of its partnership with the pilot program. The plan’s emphasis on rapid retraining programs for displaced workers and AI pre-apprenticeship programs acknowledges the mistakes of previous technological transitions and demonstrates a commitment to delivering the benefits of AI to all Americans.

These proposals complement Anthropic’s efforts to understand how AI is transforming the economy. The Economic Index and Economic Futures Program are intended to provide researchers and policymakers with the data and tools needed to ensure that the economic benefits of AI are broadly shared and its risks are appropriately managed.

Supporting Safe AI Development

Powerful AI systems will be developed in the coming years, and the plan appropriately emphasizes defending against the misuse of powerful AI models and preparing for future AI-related risks. Anthropic appreciates the administration’s prioritization of research into AI interpretability, AI control systems, and resilience against attacks. These are important lines of research that must be supported to help address powerful AI systems.

Anthropic is pleased that the plan affirms the important work of the National Institute of Standards and Technology’s Center for AI Standards and Innovation (CAISI) in evaluating frontier models for national security concerns, and looks forward to continuing to work closely with it. Anthropic recommends that the administration continue investing in CAISI. As noted in Anthropic’s response, advanced AI systems are showing concerning improvements in capabilities relevant to biological weapons development. CAISI plays a leading role in developing testing and evaluation capabilities to address these risks, and Anthropic recommends focusing these efforts on the most unique and acute national security risks that AI systems may pose.

The Need for a National Standard

In addition to testing, Anthropic believes that basic AI development transparency requirements, such as public reporting on safety testing and capability evaluations, are essential for responsible AI development. Leading AI model developers should be held to basic, publicly verifiable standards for evaluating and managing the catastrophic risks posed by their systems. Anthropic’s proposed Frontier Model Transparency Framework focuses on these risks. Anthropic would like to see the report do more in this area.

Leading labs, including Anthropic, OpenAI, and Google DeepMind, have already implemented voluntary safety frameworks, demonstrating that responsible development and innovation can coexist. With the launch of Claude Opus 4, Anthropic proactively activated ASL-3 protections to prevent misuse for the development of chemical, biological, radiological, and nuclear (CBRN) weapons. This precautionary step shows that robust safety protections not only do not slow innovation, but also help build better and more reliable systems.

Anthropic shares the administration’s concerns about overly prescriptive regulatory approaches that create an inconsistent and burdensome patchwork of laws. Ideally, these transparency requirements should come from the government in the form of a single national standard. However, consistent with Anthropic’s stated belief that a ten-year moratorium on state AI laws is too blunt an instrument, Anthropic continues to oppose proposals aimed at preventing states from taking action to protect their citizens from potential harm caused by powerful AI systems if the federal government fails to act.

Maintaining Strong Export Controls

The plan states that "denying our foreign adversaries access to [advanced AI compute] . . . is a matter of both geostrategic competition and national security." Anthropic strongly agrees. Anthropic is therefore concerned about the administration’s recent reversal regarding exports of Nvidia H20 chips to China.

AI development has been defined by scaling laws: a system’s intelligence and capabilities are determined by the scale of compute, energy, and data inputs used during training. While these scaling laws continue to hold, the latest and most capable reasoning models have shown that AI capability scales with the amount of compute available to a system working on a given task, or "inference." The amount of compute available during inference is limited by a chip’s memory bandwidth. While the H20’s raw computing power is surpassed by chips manufactured by Huawei, as Commerce Secretary Lutnick and Under Secretary Kessler recently testified, Huawei continues to struggle with production volume, and no domestically produced Chinese chip matches the H20’s memory bandwidth.

As a result, the H20 provides unique and critical computing capabilities that would otherwise be unavailable to Chinese companies and offsets what would otherwise be a major shortage of AI chips in China. Permitting H20 exports to China would squander an opportunity to extend American dominance in AI just as a new phase of competition begins. Moreover, exports of American AI chips will not deter the Chinese Communist Party from pursuing self-sufficiency across the AI stack.

Anthropic therefore strongly recommends that the administration maintain controls on the H20 chip. These controls are consistent with the export controls recommended in the Action Plan and are essential to securing and extending American leadership in AI.

Looking Ahead

The alignment between many of Anthropic’s recommendations and the AI Action Plan demonstrates a shared understanding of AI’s transformative potential and the urgent steps needed to maintain American leadership. Anthropic looks forward to working with the administration to implement these initiatives while ensuring appropriate attention to catastrophic risks and maintaining strong export controls. Together, we can ensure that powerful AI systems are developed safely in America, by American companies, reflecting American values and interests.