The U.S. Government’s New Restrictions on AI Models
Anthropic PBC has taken the step of disabling access to its most advanced artificial intelligence models, including Mythos, in response to an unprecedented directive from the Trump administration. This order aims to prevent foreign nationals from accessing the technology, regardless of their location. According to a statement from the company, the U.S. government instructed Anthropic to suspend access to the Fable 5 and Mythos 5 models for any foreign national. The Commerce Department confirmed this directive, leading Anthropic to shut off access to both systems for all customers.
This move marks a significant shift in how the U.S. government is handling frontier AI models developed by American companies. While previous administrations have imposed restrictions on other technologies like semiconductors and supercomputers, this is the first time such sweeping measures have been applied directly to AI software. These actions have sparked debates over constitutional and commercial implications.
Concerns Over AI Model Security
Anthropic expressed disagreement with the government’s decision, arguing that the finding of a narrow potential jailbreak should not warrant recalling a commercial model used by hundreds of millions of people. The company believes that applying this standard across the industry could halt all new model deployments.
According to reports, researchers at Amazon.com Inc. had identified vulnerabilities in Anthropic’s Fable 5 model. Amazon CEO Andy Jassy was involved in discussions with U.S. officials about these vulnerabilities. However, Amazon declined to provide details about these conversations, stating that it is common for governments to consult on security risks.
Implications for the AI Industry
The government’s broad restrictions on AI models could set a precedent for major AI developers such as OpenAI, Alphabet Inc.’s Google, and Meta Platforms Inc. Industry leaders like Nvidia Corp. CEO Jensen Huang and OpenAI CEO Sam Altman have previously advocated for promoting global adoption of American AI systems rather than restricting access.
Aidan Gomez, co-founder of Cohere Inc., noted that this move serves as a wake-up call for the industry. “No one can deny it any more,” he said. The directive also threatens to escalate tensions between Anthropic and the Trump administration, which had previously clashed over the use of its technology for military and surveillance purposes.
Anthropic’s Position and Challenges
Privately held Anthropic has positioned itself as a responsible AI developer. It initially released its Mythos model in April to a limited group of companies and institutions, highlighting the risks associated with its cybersecurity capabilities. Despite these precautions, the company faced scrutiny from the Trump administration, which declared it a supply-chain risk and ordered U.S. agencies to phase out its products.
David Sacks, former AI czar under Trump, claimed that Anthropic refused to address a jailbreak vulnerability in its Fable model. He stated that the administration hopes for a resolution that would allow Fable to return to general release.
The Broader Impact on AI Development
The latest restrictions are colliding with a race among U.S. developers to deliver the most advanced AI models. Both OpenAI and Anthropic are aiming for initial public offerings this year. Anthropic recently posted a blog calling for a system where governments and AI developers collectively decide when to slow work on the technology to mitigate risks.
The European Union is assessing Anthropic’s statements and continuing discussions with allies about the potential risks of powerful AI models. The European Commission emphasized the need for technological sovereignty amid these developments.
Cohere’s Gomez expressed concerns about the impact on international technological alliances, stating that this move is not partnerly and may hinder progress. As the AI landscape continues to evolve, the balance between innovation and security remains a critical challenge for developers and policymakers alike.