Trump export order pulls Anthropic’s new models offline

Trump export – Anthropic took its two newest AI models, Fable 5 and Mythos 5, offline after a Trump administration export control order. The move has triggered a backlash from cybersecurity experts, who argue the action wrongly disrupts network defenders in the U.S.—and coul

Late last Friday, Anthropic moved quickly. Its two newest AI models—Fable 5 and Mythos 5—were taken offline after an export control order from the Trump administration, according to the discussion on TechCrunch’s Equity podcast.

For listeners following the model lineup, the timing felt particularly abrupt: Fable 5 was more available to the public, while Mythos 5 was the one available to current Mythos users. Then, in the middle of a tense policy moment, both were pulled.

The letter that Anthropic received, sent last Friday, cited “national security concerns.” No details about those concerns were made public. Anthropic was told it had to ensure the models couldn’t be used by any foreign nationals—an instruction that immediately created a practical problem for the company. In the podcast discussion. Rebecca Bellan said Anthropic effectively had to take the models down entirely because it couldn’t reliably tell when users—including its own employees—might be foreign nationals.

Behind the scenes. the reason the White House acted so swiftly appears to have traced back to a finding by Amazon researchers. As the conversation described it, they allegedly found a way to bypass Fable 5’s guardrails. Amazon CEO Andy Jassy then raised the concern with the White House, and the situation “spiraled from there,” Bellan said.

Sean O’Kane said the speed of it all stood out—especially the way it moved from a Friday afternoon into the weekend. He tied the moment to a broader feeling of politics and distraction. noting that at the same time the administration was ostensibly negotiating some kind of treaty for the war it started in Iran.

The podcast also sharpened a key question: if this can happen to Anthropic, what happens to everyone else?

O’Kane pointed out that Anthropic “has not had the best relationship with the Trump administration” compared with other leading AI labs. That, he suggested, could mean other companies might not face the same kind of crackdown.

But Bellan and the guests emphasized that the justification for the export control doesn’t look clean to outside experts. They discussed how leading cybersecurity experts have signed an open letter urging Trump to revoke the order. Those experts, as Bellan described, argue it’s dangerous to pull advanced cybersecurity capabilities away from U.S. network defenders.

The friction didn’t end with the order itself. Bellan said cybersecurity researchers also argue that the kinds of issues raised—jailbreaks—are not unique to Anthropic. Anthropic’s own position, she said, is that the same jailbreaks could be found in several other AI models.

That contradiction—an order framed around national security, combined with claims that the underlying security risk isn’t uniquely Anthropic—pulled the discussion toward a sharper theme: whether this is primarily about security, or about something else.

Anthony Ha said independent security reporting and expert analysis suggest the security risk from Anthropic isn’t “that unique.” In his telling. the situation seemed less like a clear-cut technical outlier and more like a larger conflict that escalates once parts of the Trump administration and Anthropic are already at odds. He warned that even if a given risk exists. it can become distorted when relationships deteriorate to the point where everything gets blown up.

That led to a more uncomfortable tradeoff for rivals. Ha said that for another company, it might feel advantageous if Anthropic is the one the government targets. But it’s also not comforting: “It’s not a great regulatory landscape” if the fear is that the tap can get turned off for reasons that aren’t strictly tied to uniquely identifiable technical risk.

Bellan’s frustration sounded more direct. She described the action as feeling retaliatory—especially because, in her view, the government had already labeled Anthropic a supply chain risk, and a major lawsuit is ongoing between them.

And she pointed to the company’s own messaging about danger. Bellan said there was public talk from Anthropic about slowing AI because it was becoming dangerous. Then. she argued. Fable was released only about a week after those warnings—presented as “our most insane ever. super powerful model.” The contrast. she said. fueled cynicism that the company wanted it both ways: warning about risk for everyone else while positioning itself as the responsible exception.

Ha added a parallel from the broader industry. He framed the moment as a microcosm of AI debates where leaders talk about lowering the temperature while the public remains skeptical. In his description. the relationship between promise and perceived impact—jobs. power. scrutiny—has made it easier for policymakers and critics to respond harshly rather than calmly.

Even so, the podcast didn’t leave the story entirely in anger. Ha and Bellan both circled back to a possible upside for Anthropic: the company’s models can look more compelling when they’re treated as dangerous.

Ha referenced a previous “blow-up” between Anthropic and the Trump administration. saying it was “good for the company” in at least some ways—Claude downloads rose. and people who may have treated ChatGPT as the default moved to Claude as a more “responsible” alternative. sometimes described as a kind of “resistance” option. In the same way. he said. Anthropic might face stress now. but the controversy could also make the models seem even more powerful to the public.

For Bellan, the dynamic landed in a blunt line: the public, she said, tends to love the bad boy. If the message becomes “we’re so dangerous,” the models might gain attention precisely because they were forcibly pulled.

For now. what’s clear is the immediate outcome: Fable 5 and Mythos 5 are offline after a Trump administration export control order citing “national security concerns. ” a claim experts say doesn’t add up neatly—especially when the issues reportedly span beyond Anthropic. And the political question underneath has become just as urgent as the technical one: whether the next moves in AI regulation will be driven by security evidence. or by who is already in someone’s crosshairs.

Anthropic Fable 5 Mythos 5 export control order Trump administration AI models offline cybersecurity experts open letter national security concerns Andy Jassy Fable 5 guardrails