The White House imposed sweeping export controls on AI company Anthropic after a 24-hour discussion involving top officials and the company’s CEO.

The move came after concerns were raised about Anthropic’s newly released AI model, Fable, which the company had said was safe. But soon after its launch, government officials began worrying that the model’s safety protections might not be strong enough, according to Politico.

On Thursday, just two days after the release, Amazon CEO Andy Jassy reportedly raised concerns with the White House about the model’s potential to be bypassed or “jailbroken.” Amazon is an investor in Anthropic.

Officials believed the model could pose security risks, so senior White House figures first tried to convince Anthropic to voluntarily take it down. This effort lasted around 24 hours and involved several high-level officials, including Treasury Secretary Scott Bessent, White House Cyber Director Sean Cairncross and Commerce Secretary Howard Lutnick.

Other senior officials also joined parts of these discussions, including Commerce Under Secretary Jeffrey Kessler, White House Staff Secretary Will Scharf, Deputy Chief of Staff Richard Walters and policy aide Walker Barrett. They discussed the potential risks posed by the AI model.

Amodei told US officials that the security issue they were worried about was limited and specific. He argued that this was not the same as a full “jailbreak,” where someone could completely break the AI’s safety rules and use it without any restrictions, Reuters reported.

White House officials reportedly gave Anthropic just 90 minutes to take down its AI models. A company spokesperson told Politico that the order was strict and came with a deadline.

Later, in a blog post, Anthropic defended its model. The company said that no one had been able to find a “universal jailbreak”.

“We have instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cybersecurity (among others). In fact, our safeguards are so strong that many users have complained that they are overly broad,” the blog post read.

Anthropic said no one has found a “universal jailbreak” that can fully break Fable 5’s safety systems. However, the company admitted that perfect protection is not possible for any AI model and that all systems can still be tricked in limited ways.

“Export controls were a last resort after begging them for hours to work with us. This was not something we wanted to do, but our hands were tied,” the senior White House official said, as quoted by Turkiye Today.