The US prevents foreigners from using Anthropic’s most advanced AI models
Anthropic said Friday night that the US government has ordered the company to suspend all foreign nationals’ access to its Fable 5 and Mythos 5 artificial intelligence models, citing national security concerns.
The company published the order in post on a social networkstating that “they must suddenly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.” The suspension also applies to foreign nationals who work for Anthropic.
The order to restrict access to the new AI models came from the Commerce Department, the person familiar with the matter said, speaking on condition of anonymity. It was unclear how long the restrictions would last.
The alien limits represent a new chapter in Anthropic’s often prickly relationship with the federal government, and are a highly unusual example of federal authorities dictating how a tech company can operate.
It’s also the second time the Trump administration has targeted the company, which is based in San Francisco. In March, after tense talks about how Anthropic’s AI could be used in military and intelligence environments, the Pentagon said it considered Anthropic an unacceptable supply chain risk, potentially limiting its use by federal agencies.
Government restrictions on the sale of key technologies such as computer chips to countries such as China are common. But the order restricting who can use Anthropic’s technology is unusually broad and could theoretically prevent Anthropic employees who are citizens of allied nations such as Canada or Britain from working on it.
The deal could also cause significant problems for technology development at Anthropic, which, along with OpenAI, is one of the most influential AI companies in the world. Although new, Mythos and Fable 5 were considered significant improvements to earlier AI models.
Anthropic declined to comment beyond its post, and the White House did not immediately respond to a request for comment.
Anthropic unveiled Claude Mythos in April, but said it could not share the AI system with the public because it could become a powerful tool for hackers looking for ways to break into computer networks.
Mythos quickly became a security concern in Washington and Silicon Valley because it could represent what experts have feared for several years: that AI poses a new and far more dangerous threat to computer networks.
Anthropic has shared the Claude AI system with about 40 organizations that maintain critical computing infrastructure so they can use the system to patch security vulnerabilities before hackers exploit them.
Reactions from researchers who had access to the Mythos varied. Some agreed that the Mythos was a new and worrisome threat. But others have argued that it’s more evolutionary than revolutionary, and that they can use it to protect their networks, just as hackers can use it to break into them.
In a brief statement on Friday, Anthropic appeared to indicate that it viewed the government’s decision as an overreaction. “We apologize for this disruption to our customers,” the company said. “We believe this is a misunderstanding and are working to restore access as quickly as possible.”
After Anthropic exposed Mythos, the White House began considering new oversight of AI systems. Last week, in a major policy shift for an administration that has been very cautious about AI, President Trump signed an executive order asking companies to voluntarily provide government oversight of new systems before releasing them to the public.
Earlier this month, Anthropic said it would share the system with about 150 other organizations. Then on Tuesday, the company unveiled a straitjacket version of the Mythos called the Claude Fable 5.
The Fable system, the company said, included additional guardrails designed to block responses in areas such as cybersecurity, biology and other sensitive topics. Railing can prevent malicious hackers from using the technology to attack computer networks. But the railing can also prevent companies and cybersecurity professionals from using the system to defend against attacks.
Most of the queries it considers risky would instead be transferred to an older AI system called Opus 4.8, according to Antropic.
Anthropic’s Opus 4.8 and other older AI models were not affected by Friday’s government order, the company said.
Cybersecurity experts disagreed on whether Anthropic was right to limit the release of Mythos, pointing out that this kind of system could be used to both attack and defend computer networks. Other companies, notably Anthropic rival OpenAI, offer similar technology.
Former US cybersecurity and technology policy officials said they were surprised by the news, calling it highly unusual and at odds with other moves by the Trump administration, such as allowing the sale of some advanced AI chips to China.
“I’m speechless,” Dean Ball, a senior fellow at the Foundation for American Innovation and a former AI adviser to Mr. Trump, said in a social media post. He added that the decision was “confused”.
A week after Anthropic revealed Mythos, OpenAI said it would also scale back the release of its latest system, called GPT-Cyber 5.4. But OpenAI shared its technology with a larger group, initially sharing the system with hundreds of experts before opening it up to thousands.