US Removes Restrictions on Anthropic Claude Models After Security Talks With White House

The Trump administration has removed export controls on Anthropic’s most advanced AI systems, reversing restrictions that had temporarily limited access to the company’s frontier models after national security concerns surfaced earlier this month.

The move reopens the more advanced models of Anthropic, some of which had been placed under tight restrictions limiting access outside the U.S. and requiring closer monitoring of where they are used.

The move comes after weeks of negotiations between Anthropic executives and the U.S. Department of Commerce to resolve cybersecurity and model safety risks that prompted the original intervention, officials said.

The export controls were first implemented after federal officials said Anthropic’s most powerful models could be vulnerable to misuse, including potential cyber exploitation or disclosure of sensitive government systems.

Those restrictions effectively halted or limited international deployment of some models, disrupting enterprise customers and developers that depend on Anthropic’s AI infrastructure.

The administration at the time argued that frontier AI systems needed stronger oversight due to their rapidly developing capabilities and potential for dual-use applications for both commercial and national security purposes.

As part of negotiations with U.S. authorities, Anthropic agreed to implement tighter safety measures and monitoring protocols This includes wider incident reporting requirements and closer coordination with federal agencies to identify possible misuse scenarios.

The company also pledged to implement strengthened safeguards against unauthorized access and to reduce cybersecurity risks connected with model deployment.

The Commerce Department then acted to undo the export caps and restore wider access to the models.

Looser controls are expected to relieve pressure on global customers who rely on Anthropic’s models for enterprise AI applications, such as software development, research and automation systems.

The incident underscores mounting strains between fast-moving AI progress and government attempts to regulate frontier systems in real time, industry analysts say. The back and forth also underscores the growing power of national security agencies to determine the release and deployment of advanced AI models.

Access has been restored, but officials warned that new restrictions could be imposed if they see new risks, a sign of a more fluid regulatory approach to AI oversight.

The dispute is another episode in the shifting policy framework for advanced artificial intelligence in the U.S. As AI capabilities expand, policymakers are still trying to find the right balance between innovation, commercial competitiveness and security risk.

Anthropic has become an influential voice in AI safety research and its interaction with regulators is part of a broader industry pivot toward greater collaboration with governments on frontier model governance.

The easing of export controls indicates a momentary thaw in tensions, but the larger question of how to control powerful AI systems remains.