U.S. restrictions on new Anthropic models could trigger a global AI arms race

By Emily Bary

Anthropic has blocked access to its Fable 5 and Mythos 5 models after the government flagged them as a security threat when used by foreign nationals

Anthropic has had to yank down Mythos and another AI model after government pushback.

Training artificial-intelligence models is expensive, and the big AI labs have been banking on paid usage to help justify their skyrocketing costs. A fresh development involving Anthropic suggests that the biggest corporate developers may face even more roadblocks going forward.

It's already been challenging for AI companies to effectively monetize their services; now Anthropic has found itself in the U.S. government's crosshairs once more. The company disclosed late Friday that it had to "abruptly disable Fable 5 and Mythos 5," two of its new models, due to export-control restrictions issued by the U.S. government.

The move sets the stage for AI to be increasingly regulated for political and geopolitical purposes, raising the stakes for countries to develop their own AI models so that they're not denied access to major private ones going forward.

"AI is more powerful than nuclear weapons," the founder of Switzerland-based Edelbridge Capital said on X. "Every country who will not secure their access to frontier AI will be left behind."

That's potentially good news for suppliers of AI infrastructure, which could see heightened demand for AI hardware and related products if countries feel a greater urgency to build AI models of their own. But it also stands to make the market more competitive for big players like Anthropic and OpenAI that need to ramp up paid usage to solidify the economics of AI development.

Before the models were released, Anthropic warned that Mythos was so powerful that it could potentially be dangerous in the hands of bad actors. Anthropic delayed the release and assembled a coalition of major companies that tested the model before it made its public debut.

See also: How Elon Musk nailed the SpaceX IPO

The U.S. government directive was to "suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States," according to an Anthropic blog post on Friday evening. The rule even applies to foreign nationals who are Anthropic employees themselves. But on a practical level, it appears difficult to only block foreigners from using the two models - hence the decision to cut access for all users.

The Wall Street Journal reported that Amazon researchers flagged the ability for Anthropic's new models to be jailbroken, or made to reveal information about security vulnerabilities.

"As a leading cloud provider that serves a large number of private- and public-sector customers, it's not uncommon for governments to seek our counsel on potential security risks," an Amazon spokesperson said in a statement to MarketWatch. "When they occur, we don't share the details of these discussions."

When reached for comment, Anthropic referred back to its blog post as its official statement.

"We suspect that perfect jailbreak resistance is not currently possible for any model provider," that blog post noted.

Read: Will AI start 'going rogue'? The chorus of warnings is getting louder.

The U.S. government's actions struck some experts as misguided. Chris McGuire, a senior fellow with the Council on Foreign Relations, noted on X that the administration is broadly blocking Mythos 5 and Fable 5 access while still allowing the sale of advanced U.S. chips - which help train the smartest AI models - to Chinese companies.

"We urgently need a smart export-control strategy that applies robust export controls to deny our adversaries access to advanced technology, while advantaging U.S. companies," McGuire wrote. Recent government moves "are consistently doing the opposite" in his view, he added.

See also: Here's how far the Trump administration's 'startling turn' on AI regulation might go

-Emily Bary

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

06-13-26 1411ET