In 24 hours, White House exports controls hit Anthropic’s Fable 5 and Mythos 5

The Trump administration imposed export controls on Anthropic’s latest models, Fable 5 and Mythos 5, on Friday. The measures forced Anthropic to disable access for foreign nationals and effectively “abruptly disable” the models for all customers to ensure compliance, according to Anthropic. The whole chain of events was triggered by a frantic, 24-hour effort to get Anthropic to voluntarily pull the newly released model over what officials believed were security risks.

At the center of that rush was Anthropic CEO Dario Amodei, who was ordered to block foreign access. Administration officials and a senior White House official, granted anonymity because they were not authorized to discuss the episode publicly, described tense calls between Amodei and senior government figures including Treasury Secretary Scott Bessent, White House Cyber Director Sean Cairncross, and Commerce Secretary Howard Lutnick. Anthropic later disputed that Amodei had been unavailable due to a wellness retreat, and it also said in a post after the export controls that the move was “disproportionate.”

So what actually happened in that 24-hour window? According to the administration officials and the senior White House official, the export controls were preceded by multiple tense calls and an attempt to persuade Anthropic to voluntarily remove a model released to the public. Officials said they believed the model posed security risks, specifically pointing to concerns that the model’s guardrails could be bypassed. One day earlier, Amazon CEO Andy Jassy raised those guardrail concerns with the White House on Thursday, two days after Fable’s public release, according to the officials. Amazon is an investor in Anthropic, and a person familiar with Amazon’s discussions said Amazon was responding to an administration request for feedback.

By Friday morning, the issue had reportedly escalated to the highest levels. Bessent, Cairncross, Chief of Staff Susie Wiles, and other senior officials met to discuss the model and the administration’s response, with Bessent joining remotely while traveling to Houston for a scheduled public event. The administration then attempted to reach Amodei but was told he was unavailable because he was at a wellness retreat, according to an administration official and the senior White House official. Anthropic’s spokesperson rejected that claim and called it “absolutely false.” A person close to Anthropic said Amodei was first requested around noon and was on the phone with senior officials within an hour and 15 minutes, and that when he was out of pocket Anthropic offered other senior leaders to participate.

Once they reached him, Amodei participated in three calls that involved roughly half a dozen senior officials, including Cairncross, Bessent, and Lutnick, according to the senior White House official and one of the administration officials. In those calls, the reporting says Amodei pushed back, defended the guardrails, and argued that the specific bypass type officials had identified did not pose the same risk as a broader “jailbreak” that would allow use without any guardrails. That argument is consistent with Anthropic’s later explanation after export controls were put in place. In a blog post, Anthropic said that “no testers have yet been able to find a universal jailbreak - a jailbreak method that can very broadly bypass the model's safeguards, unblocking a wide range of cyber capabilities,” adding that total avoidance of jailbreaks is not now possible for them or any other companies.

But officials were not convinced. A White House official said that Amazon’s findings were run past the National Security Agency and that officials felt they had “proof.” They urged Anthropic to voluntarily remove the model and coordinate with the government to address the vulnerabilities, according to the senior White House official and the two administration officials. Amodei asked for more time and information, but the reporting says he made no commitments to pull the model. At one point, Bessent told Amodei directly that he was making a “bad decision.” Shortly after the calls, the Trump administration imposed the export controls on Fable 5 and Mythos 5, citing national security authority and banning their use by foreign nationals. Anthropic said the “net effect” was that the order “abruptly disable[d]” the models for all customers to ensure compliance.

Behind the rhetoric is a key operational reality: export controls change the governance problem from “risk assessment and iterative fixes” into “instant compliance or shutdown.” Anthropic’s post after the controls argued that it was complying with the government directive but called the action disproportionate, saying it believes the government should have the ability to block unsafe deployments through a statutory process that is transparent, fair, clear, and grounded in technical facts. Meanwhile, a White House official, speaking anonymously, said innovation remains the White House’s “number one goal,” but it also has to prioritize security.

The strategic stakes extend beyond Anthropic and Amazon. Anthropic has positioned itself as a vocal advocate for AI regulation to counter global security risks and job disruption as AI advances quickly. Yet this episode shows how regulation can arrive with an emergency brake: senior officials tried to get a voluntary pull first, then moved to export controls within hours once the effort did not succeed. Three people familiar with the government’s thinking said Amazon wasn’t the only company raising concerns and that the “crux of the issue” was the lack of seriousness Anthropic was applying to the risk. A second person close to Anthropic rejected the idea that the “jailbreak” was a breakdown of Fable 5’s safety systems, pointing to the company’s collaboration with the administration before release.

For other founders, CEOs, and boards watching from the outside, the non-obvious lesson is about process control, not just model safety. If you are selling frontier AI to the world, you are also effectively participating in national security triage, where a decision can hinge on speed, perceived evidence, and whether leaders treat a reported bypass as a must-fix vulnerability or as something narrower than officials believe. And once export controls hit, the blast radius can be immediate, not incremental. For decision-makers, that’s the part to underline: the “whirlwind” ended with a shutdown that even the company framed as abrupt, and the policy direction now visibly rewards responsiveness under pressure, not arguments after the deadline.