US restricts Anthropic AI models over national security concerns

The US government has ordered artificial intelligence company Anthropic to halt access to some of its most advanced AI models.

The US government has ordered artificial intelligence company Anthropic to halt access to some of its most advanced AI models after officials raised concerns that the technology could be exploited by foreign military and intelligence agencies.

US Commerce Secretary Howard Lutnick informed Anthropic in a letter that the company’s recently released Mythos and Fable AI models must no longer be made available to users outside the United States or to foreign nationals. The decision was driven by fears that the technology could eventually be used by intelligence services or military organizations in countries viewed by Washington as strategic rivals, including China and Russia.

The move has triggered urgent discussions between Anthropic and the US government. Senior technical representatives from the company met Commerce Department officials in Washington to seek a resolution that would allow access to be restored while addressing national security concerns. Talks are expected to continue as both sides attempt to find a compromise.

At the center of the dispute is concern over the cybersecurity capabilities of Anthropic’s newest AI systems. Government officials reportedly believe there may be ways to bypass safety protections built into the models. Such vulnerabilities could potentially allow users to identify software weaknesses or assist in cyber operations. Anthropic has acknowledged that researchers discovered methods of circumventing certain safeguards but argues that the issues were limited in scope and similar capabilities already exist in other publicly available AI systems.

Following the government’s directive, Anthropic disabled access to the affected models worldwide. The company said it took the action to comply with federal instructions while continuing discussions with regulators. Company officials maintain that security protections were thoroughly tested before release and that government agencies had reviewed the technology during development.

The dispute highlights growing tensions between national security priorities and the rapid advancement of artificial intelligence. The Trump administration has increasingly focused on controlling the spread of powerful AI technologies, particularly those that could have military, intelligence or cybersecurity applications. Officials are seeking stronger guarantees that advanced AI systems cannot be used in ways that threaten US interests.

Relations between Anthropic and the administration have reportedly been strained for several months. Earlier disagreements emerged after the company declined to allow its AI models to be used for certain military applications, including domestic surveillance and fully autonomous weapons systems. Those disputes contributed to a broader debate over the role of private technology firms in national security programs.

The government’s latest action is notable because it represents the first reported use of powers granted under the Export Control Reform Act of 2018 to restrict access to advanced AI models. The Commerce Department has warned that any violation of the order could result in significant civil and criminal penalties.

However, legal experts have questioned whether existing export control laws are fully suited to regulating artificial intelligence. Unlike traditional technologies, AI models are often accessed remotely through online services rather than being physically exported across borders. This has raised questions about how such restrictions can be enforced and whether current regulations provide sufficient legal authority.

The decision has also sparked opposition from parts of the technology and cybersecurity community. More than 80 industry leaders and security experts have urged the administration to reconsider the restrictions, arguing that limiting access could hinder innovation and damage the competitiveness of American AI companies.

As negotiations continue, the outcome of the dispute may shape how governments regulate advanced artificial intelligence in the future and determine the balance between technological progress and national security concerns.