Artificial intelligence company Anthropic has restricted access to two of its latest AI models after receiving an order from the United States government citing national security concerns.

In a blog post published on Friday, the company said federal authorities instructed it to prevent foreign nationals from accessing its newly released Fable 5 and Mythos 5 models. According to Anthropic, the directive was delivered at 5:21 p.m. local time on Friday and did not include detailed information regarding the specific security concerns that prompted the action.

Restrictions Extend to Foreign Nationals in the United States

Anthropic said the order applies not only to users outside the United States but also to foreign nationals currently residing or working within the country, including employees of the company.

As a result, the firm said it was required to suspend access for affected users on short notice while it works to understand and comply with the government's requirements.

Advanced Cybersecurity Capabilities Draw Scrutiny

The company's Mythos 5 model is designed with advanced capabilities for identifying and analyzing software vulnerabilities, including flaws that may have remained undiscovered for extended periods. Anthropic said the technology has been used by government agencies and selected private-sector partners to strengthen cybersecurity defenses and address software security weaknesses.

At the same time, experts have long debated the potential risks associated with increasingly powerful AI systems, particularly concerns that advanced cybersecurity capabilities could be misused for offensive cyber operations if deployed irresponsibly.

Fable 5 Released With Safeguards

Anthropic said its newly launched Fable 5 model is built on the same underlying technology as Mythos 5 but includes restrictions on certain cybersecurity and biotechnology-related functions.

Mythos 5, the company's non-public flagship model, remains available only to government agencies and a limited group of corporate partners for authorized security and research applications.

Company Seeks Greater Clarity

Anthropic said it has received only partial information regarding the government's decision and is continuing discussions with officials to better understand the basis for the restrictions.

The company added that it reviewed a report it believes may have contributed to the government's concerns. According to Anthropic's assessment, the report focused on the model's ability to analyze software code, identify errors, and suggest corrections.

Anthropic noted that similar capabilities are available in AI systems offered by other developers, including models produced by OpenAI and other major technology companies.

Anthropic Disputes Basis for Ban

The company said it disagrees with the view that such capabilities justify restricting access to software used by large numbers of people. Anthropic stated that Fable 5 underwent extensive safety testing prior to its release and includes safeguards intended to limit misuse.

The company emphasized that it remains committed to cooperating with regulators while continuing to develop and deploy AI systems responsibly.

Broader Debate Over Advanced AI Development

The restrictions come amid growing debate over the pace of artificial intelligence development and the potential risks associated with increasingly capable systems.

Earlier this month, Anthropic called for greater coordination among leading AI developers and suggested that the industry should be prepared to slow or temporarily pause the development of the most advanced models if necessary.

In a separate blog post, the company argued that rapid advances in AI capabilities could eventually outpace existing governance and safety mechanisms, making it important for governments and technology firms to retain the ability to moderate the pace of development when warranted.