Anthropic discontinues cutting-edge AI models including Mythos after US bans access from abroad

issued Sunday, June 14, 2026 · 09:51 AM

ANTHROPIC has revoked access to its most advanced artificial intelligence models, including Mythos, following an unprecedented order by the Trump administration to keep the technology out of the hands of all foreigners.

The U.S. government has directed Anthropic to stop access to its Fable 5 and Mythos 5 models by foreign nationals “both within and outside the United States,” citing national security concerns, the company said in a statement. U.S. officials confirmed the Commerce Department sent the letter. The model developer then blocked access to both systems by all customers to ensure compliance.

Never before has the U.S. government taken such drastic steps to curtail foreign access to frontier AI models developed by U.S. companies. The Trump and Biden administrations have restricted foreign access to other critical technologies such as semiconductors and supercomputers, and some have debated the merits of blocking access to AI models. However, restrictions on the software itself raise constitutional and commercial concerns.

Anthropic said it believed the U.S. government issued the order after discovering that it was possible to “jailbreak” Fable 5, a recently released version of Mythos that blocked the company from performing cybersecurity operations, or bypass guardrails.

“We do not agree that the discovery of a narrow jailbreak possibility should be cause for a recall of a commercial model that has been deployed to hundreds of millions of people,” Anthropic said in a post on its website. “If this standard were applied industry-wide, we believe it would effectively halt the rollout of all new models to all Frontier model providers.”

Anthropic model vulnerabilities

According to a report, Amazon.com researchers conducted jailbreak research and uncovered several vulnerabilities in Anthropic’s model. wall street journal.

Amazon and the U.S. government had been in touch about the vulnerability before the restrictions were imposed, said the people, who were granted anonymity to discuss sensitive matters. Amazon CEO Andy Jassy was involved in the conversations, one of the people said. The Information earlier reported that Mr. Jassy had expressed his concerns to senior U.S. officials.

An Amazon spokesperson said it is not unusual for governments to consult with the company about security risks, but declined to provide details of such discussions.

The government’s move to broadly restrict access to a range of AI models in the name of national security threatens to set a precedent for all major AI model developers, including OpenAI, Alphabet’s Google, and Metaplatform. Industry leaders such as Nvidia CEO Jensen Huang and OpenAI CEO Sam Altman have in the past encouraged the U.S. government to instead promote global adoption of U.S. AI systems and protect the country’s leadership.

“This is a big wake-up call for people who were naive and perhaps hoped this influence would not be exercised,” Aidan Gomez, co-founder of NVIDIA-backed AI startup Cohere, said in an interview on Saturday (June 13). “No one can deny that anymore.”

Anthropic announced Friday that it had received the government order at 5:21 p.m. New York time. The termination order contradicts previous statements and a recent executive order signed by President Donald Trump, indicating that the administration will not pursue a licensing system to review models.

Friday’s directive also threatens to inflame long-standing tensions between Anthrop and some within the Trump administration. Earlier this year, the AI ​​developer clashed with the Pentagon over the use of the technology for military and surveillance purposes. The administration declared the company a risk to the U.S. supply chain as a result of the explosion and ordered U.S. government agencies to phase out the use of the company’s products.

Too risky to distribute more widely

Privately held Anthropic, which has long positioned itself as a more responsible AI developer, first released the Mythos model to a very narrow group of companies and institutions in April, warning that its ability to discover cybersecurity vulnerabilities made it too risky for wider distribution.

There were signs that the limited release was working to ease tensions between Anthropic and the Trump administration. In April, the U.S. government was preparing to make a version of Mythos available to key federal agencies, Bloomberg previously reported.

Mythos also accelerated the Trump administration’s AI policy efforts, including a recent executive order calling for a voluntary model review. The order clearly states that nothing in it should be construed as creating a compulsory licensing regime.

David Sachs, President Trump’s former AI czar and current co-chair of the President’s Science and Technology Advisory Council, said Anthropic refused to fix the guardrail jailbreak in its Fable model.

“The administrator’s hope now is that Anthropic improves its safety issues, export restrictions are lifted, and Fable returns to public release,” he wrote in a post on X. “Administrators want all of this to happen as soon as possible. We are frankly perplexed by Anthropic’s reluctance to comply with safety requests, which they have previously said are their top priorities.”

Recent government restrictions have clashed with a race among U.S. AI developers to offer cutting-edge AI models and prove to investors that their technology can be profitable. Both OpenAI and Anthropic are aiming for initial public offerings as early as 2026, following SpaceX’s own historic IPO.

In the rush to deliver state-of-the-art AI models, Anthropic itself posted a lengthy blog in early June calling for a system in which governments and AI developers can jointly decide when to slow down work on a technology to avoid the risks it may pose.

“The world would be better off if we had the option to view or pause” potentially dangerous AI work, the company said in a post at the time. AI has advanced to the point where it can make human tasks thousands of times more efficient or even replace them, creating new risks, the company said.

The European Union’s executive arm said it was assessing Anthropic’s statement and was continuing to consult with allies about the potential risks and cybersecurity concerns associated with powerful new AI models. The European Commission added that recent developments highlight the need for European technological sovereignty.

“As someone on the ground, I’m not particularly excited to see this. I don’t think this is a partnership. I don’t think this is the right move for the broad technology partnership that has developed over the last 80 years,” Kohia’s Gomez said. bloomberg