Anthropic Response to America's AI Action Plan

Anthropic has expressed support for the White House's "Winning the Race: America's AI Action Plan," noting that many of the administration's strategies align with Anthropic's own policy recommendations. While the company endorses the plan's focus on infrastructure and federal adoption, it argues that strict export controls and national transparency standards are necessary to fully secure American AI leadership.

Accelerating AI Infrastructure and Federal Adoption

Anthropic supports the administration's commitment to streamlining energy and data center permitting to meet the power demands of AI development. The company warns that insufficient domestic energy capacity could force American AI developers to move operations overseas, potentially exposing sensitive technology to foreign adversaries.

To increase federal AI adoption, Anthropic highlights several key proposals within the Action Plan that align with its priorities:

Resource and Procurement Reform: Tasking the Office of Management and Budget (OMB) to resolve resource constraints and procurement limitations.
Regulatory Reform: Launching a Request for Information (RFI) to identify and reform federal regulations that hinder AI innovation.
Deployment Standards: Updating federal procurement standards to remove barriers to AI system deployment.
National Security: Promoting AI adoption in defense and national security through public-private partnerships.

Democratizing AI Benefits and Research

Anthropic endorses the continuation of the National AI Research Resource (NAIRR) pilot, which provides students and researchers with the resources needed to contribute to the AI frontier. The company also supports the Action Plan's emphasis on retraining programs for displaced workers and pre-apprenticeship AI programs to ensure the economic benefits of AI are broadly shared.

To support these goals, Anthropic utilizes its Economic Index and Economic Futures Program to provide data and tools for researchers and policymakers to manage AI's economic risks and benefits.

Secure AI Development and Safety Research

Anthropic commends the administration's focus on defending against the misuse of powerful AI models and its support for research into AI interpretability, adversarial robustness, and AI control systems.

Specifically, the company supports the continued investment in the National Institute of Standards and Technology's Center for AI Standards and Innovation (CAISI). Anthropic notes that advanced AI systems are showing concerning improvements in capabilities related to biological weapons development, and CAISI's role in developing testing and evaluation capabilities for these acute national security risks is critical.

The Case for National Transparency Standards

Anthropic argues that voluntary safety frameworks are insufficient and that the government should establish a single national standard for AI development transparency. This standard would require leading developers to publicly report on safety testing and capability assessments to manage catastrophic risks.

While Anthropic supports a national standard to avoid a "burdensome patchwork of laws," it opposes a ten-year moratorium on state AI laws. The company believes states should retain the ability to protect citizens from potential harms if the federal government fails to act.

Maintaining Export Controls on H20 Chips

Anthropic strongly urges the administration to maintain export controls on Nvidia H20 chips to China. The company bases this argument on scaling laws, noting that while raw computing power may be exceeded by some Chinese chips, the H20's memory bandwidth is a critical capability that no domestically-produced Chinese chip currently matches.

Because memory bandwidth limits the compute available during inference—which is essential for the newest reasoning models—Anthropic argues that exporting H20 chips would provide Chinese firms with unique capabilities that would otherwise be unavailable, thereby undermining American AI dominance.