Newsletters
The Trump administration gave Anthropic just 90 minutes on Friday to pull its newly released artificial intelligence models before imposing sweeping export controls.
The move set off a confrontation between government security officials and the AI company, ending with two flagship models being disabled for hundreds of millions of users, according to reporting by Axios and Politico.
The controls targeted Anthropic's Fable 5 and Mythos 5 models, which the U.S. Commerce Department moved to restrict after senior White House officials concluded that a newly disclosed security vulnerability posed a national security threat, a finding Anthropic disputed, saying the issue was minor, already known, and replicated by publicly available competing models.
The sequence began Thursday, two days after Fable 5's public launch, when Amazon CEO Andy Jassy raised concerns with the White House about the model's potential to be bypassed, according to Politico, which cited two administration officials and a senior White House official who spoke anonymously.
Amazon, an investor in Anthropic, was responding to an administration request for feedback, a person familiar with Amazon's discussions told Politico.
By Friday morning, the matter had reached the highest levels of the administration.
U.S. Treasury Secretary Scott Bessent, White House Cyber Director Sean Cairncross, Chief of Staff Susie Wiles, and other senior officials convened to discuss a response. Bessent joined the meeting remotely while traveling to Houston, Politico reported.
The administration then attempted to reach Anthropic CEO, Dario Amodei.
A senior White House official told Politico that Amodei was unavailable because he was attending a wellness retreat, a claim an Anthropic spokesperson flatly rejected.
"This is absolutely false," the spokesperson said.
A person close to Anthropic told Politico that Amodei was first contacted around noon and was on a call with senior officials within an hour and 15 minutes and that Anthropic offered other senior leaders in his place while he was unavailable.
Amodei ultimately participated in three calls with roughly half a dozen senior administration officials, including Cairncross, Bessent, and Commerce Secretary Howard Lutnick, according to Politico.
Other participants included Under Secretary of Commerce for Industry and Security Jeffrey Kessler; White House staff secretary Will Scharf, White House deputy chief of staff Richard Walters; and assistant to the president for policy Walker Barrett.
During the calls, Amodei pushed back on the administration's concerns, arguing that the identified bypass was specific and limited, not a broad jailbreak that would strip the model of its protections.
Cairncross and Bessent were unmoved.
At one point, Bessent told Amodei directly that he was making a "bad decision," Politico reported. Amodei asked for more time and information but did not commit to pull the model.
A source familiar with the matter told Axios that government officials contacted Anthropic at 1 p.m. ET Friday, cited a national security threat without providing details, and ordered the models withdrawn.
By 5:21 p.m. ET, the Commerce Department sent a letter imposing export controls that restrict the use of the models by foreign nationals, inside or outside the United States, including Anthropic's own foreign-national employees.
Anthropic said it began disabling the systems upon receipt of the order.
"Export controls were a last resort after begging them for hours to work with us," the senior White House official told Politico, adding that, "This was not something we wanted to do, but our hands were tied."
A person close to Anthropic disputed that characterization, saying, "The White House gave 90 minutes to take the models down, with no details on the actual threat."
"There was never any begging or asking for them to work with us, just a declared 90-minute deadline," the person added.
In a blog post following the export control order, Anthropic said the specific bypass method the government identified involved "asking the model to read a specific codebase and fix any software flaws," a capability it said was "widely available from other models (including OpenAI's GPT-5.5) and is used every day by the defenders who keep systems safe."
The company said it had spent thousands of hours testing Fable 5 alongside the U.S. government, the United Kingdom's AI Safety Institute, and other groups before its release and that no tester had found a universal jailbreak.
"We suspect that perfect jailbreak resistance is not currently possible for any model provider," Anthropic said.
Anthropic said it would comply with the directive but called it disproportionate, warning that applying the same standard across the industry "would essentially halt all new model deployments for all frontier model providers."
The company described the situation as the result of "a misunderstanding" and said it was working to restore access as quickly as possible.
On Saturday, David Sacks, former White House AI czar and a longtime critic of Anthropic's regulatory advocacy, posted on X that he agreed with the export control decision but framed it as resolvable.
"The Admin's hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release," Sacks wrote.
"The Admin wants all of this to happen as soon as possible. It is frankly bewildering that Anthropic hasn't wanted to comply with safety requests that it previously said were its highest priority," he said.
Sacks said the decision was separate from prior administration tensions with Anthropic, which included the Pentagon designating the company a supply chain risk in March over its refusal to allow its tools to be used for mass domestic surveillance and autonomous weapons.
"The Admin values Anthropic's technical capabilities and feels that this issue, while serious, should be easily resolved," Sacks wrote.
"The ball is in Anthropic's court," he added.
The Information separately reported that U.S. officials were not expected to broaden the scope of the Anthropic-related export restrictions to additional AI companies.
A White House official said innovation remains the administration's primary goal, saying, "We also have to prioritize security as well."
Set as preferred source
Set as preferred source
5 min read
