U. S. Gov’t Forces Shutdown of Anthropic’s Newest AI Models

Export control order followed reported jailbreak of cybersecurity safeguards

The U.S. Commerce Department ordered Anthropic to cut off foreign access to its two most advanced artificial intelligence models last week, forcing the company to shut both systems down worldwide within days of their public release.

Anthropic released Claude Fable 5 on June 9. The model ran on the same underlying technology as Mythos 5, the company’s most capable system, but included extra safety restrictions meant for public release. Mythos 5 itself was shared only with a limited group of partners.

Within days, researchers at Amazon found a way around those restrictions, according to multiple news reports. Amazon is one of Anthropic’s largest investors and a major cloud computing partner.

Amazon Flags the Issue

Amazon CEO Andy Jassy raised the findings with Treasury Secretary Scott Bessent and other administration officials, according to reporting from The Wall Street Journal and other outlets. Amazon researchers reportedly used a series of prompts to get Fable 5 to produce information that could assist with cyberattacks.

An Amazon spokesperson did not confirm details of those conversations. The spokesperson said it is “not uncommon for governments to seek our counsel on potential security risks” and that the company does not discuss the specifics of those discussions.

David Sacks, a former White House AI adviser who now co-chairs the President’s Council of Advisors on Science and Technology, described the discovery in a social media post over the weekend. He said a “highly credible trusted partner of both Anthropic and the USG” had identified the workaround while testing Fable 5.

A Disputed Account

According to Sacks, the administration asked Anthropic CEO Dario Amodei to fix the flaw or pull the model from public release. Sacks said Amodei declined to do either.

Anthropic has offered a different version of events. The company said it reviewed the workaround and found it exposed only “a small number of previously known, minor vulnerabilities.” It said the same weaknesses can be found in other publicly available AI models, including OpenAI’s GPT-5.5, which are not subject to similar restrictions.

“We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” Anthropic said in a public statement.

Sacks pushed back on that characterization, arguing that a workaround capable of exposing cyberweapon-level abilities is hard to describe as anything other than serious.

Commerce Department Steps In

By Friday evening, June 12, the Commerce Department issued an export control directive barring foreign nationals from accessing Fable 5 and Mythos 5. The order applied broadly. It covered people outside the United States, foreign nationals living in the country, and even Anthropic’s own foreign-born employees.

Anthropic said it received the directive at 5:21 P.M. Eastern time. The company said the notice did not lay out specific national security concerns.

Anthropic said it had no way to separate foreign users from domestic users in real time. As a result, the company shut both models down for every customer, everywhere, to comply with the order.

“We apologize for this disruption to our customers,” Anthropic wrote in a statement posted online. “We believe this is a misunderstanding and are working to restore access as soon as possible.”

What Still Works

Anthropic said the shutdown does not affect its other Claude models, including the Opus and Sonnet lines used by most consumers and businesses. Only Fable 5 and Mythos 5 are offline.

As of this week, both models remain unavailable. Anthropic has not announced a timeline for restoring access, and the dispute between the company and federal officials remains unresolved.

The episode marks one of the most significant federal actions taken against a U.S. artificial intelligence company to date. It also highlights an unusual dynamic: Amazon, one of Anthropic’s biggest financial backers, was the party that brought the security concerns to federal officials.

Sacks said the administration acted “reluctantly” and wants the restriction lifted once the underlying issue is resolved. “The ball is in Anthropic’s court,” he wrote.

Anthropic, for its part, has said a broader standard requiring companies to pull a model over a narrow security flaw would effectively halt new AI releases across the industry.

Why it Matters in Hopkinton

AI isn’t just a buzzword in tech circles—whether you like it or not, it’s popping up in classrooms, town offices, local businesses, and in the daily lives of people all over Hopkinton.

This recent dispute really highlights something we can’t ignore: technology is racing ahead while the rules, security measures, and general understanding struggle to catch up. AI has its uses—absolutely.

But it’s not flawless or risk-free, and it’s nowhere near done evolving. Decisions about privacy, cybersecurity, how schools and businesses use AI, and whether people trust these tools aren’t just big-city concerns you see in the news. They hit home. What we decide now will shape how AI shows up, right here in Hopkinton.

This is a developing story. HopNews will provide updates as the situation surrounding Fable 5 and Mythos 5 develops.

Editor’s Note

This article was verified against multiple news sources, including Anthropic’s own public statement, reporting from Fortune, TechCrunch, Fox Business, and other outlets. Readers should treat the jailbreak’s underlying technical severity as a contested point between the company and federal officials. It’s not a settled fact.