The United States government has ordered AI company Anthropic to suspend global access to its latest artificial intelligence models, Mythos and Fable, citing concerns that the technology could be exploited by foreign military and intelligence agencies.
US Commerce Secretary Howard Lutnick directed the company to halt exports of the models to all foreign destinations and foreign nationals worldwide, according to a letter sent to Anthropic CEO Dario Amodei. The decision was driven by fears that the AI systems could be used by military intelligence organisations in countries such as China and Russia.
The move has triggered intensive negotiations between Anthropic and US officials. Senior company engineers met with representatives from the Department of Commerce in Washington to discuss possible solutions, while discussions between both sides have reportedly continued on a near-daily basis.
The dispute centres on concerns that one of Anthropic’s newest models, Fable 5, may be vulnerable to a “jailbreak” technique that could bypass built-in safeguards and enable users to identify software vulnerabilities. Anthropic acknowledged the issue but stated that the method uncovered only “minor” security flaws and that similar capabilities are already available through other publicly accessible AI models.
Following the government’s directive, Anthropic disabled access to the affected models worldwide. The company maintains that it worked closely with US authorities during the testing phase and received approval before releasing Fable 5 publicly on June 9.
Tensions between the Trump administration and Anthropic have been building for months. Relations reportedly deteriorated after the company declined to allow its AI systems to be used for domestic surveillance and fully autonomous weapons programmes. The administration subsequently placed the company on a national security blacklist.
The restrictions were imposed under powers granted by the 2018 Export Control Reform Act, which allows the US government to regulate emerging technologies considered vital to national security. Export control experts noted that this is the first known instance of the Commerce Department using those authorities against an AI company.
However, legal experts have questioned the government’s approach, arguing that AI models are generally accessed remotely rather than exported in the traditional sense. This has raised doubts about whether existing export control laws can be applied to online AI services.
The controversy has sparked significant opposition within the technology and cybersecurity sectors. More than 80 cybersecurity executives and experts signed an open letter urging the administration to reverse the restrictions. The signatories, including leaders from companies such as Nvidia and Adobe, argued that the measures could hinder innovation and undermine the cybersecurity community’s ability to develop advanced defensive tools.
As negotiations continue, both government officials and Anthropic are seeking a compromise that addresses national security concerns while restoring access to some of the company’s most advanced AI systems.
