U.S. blocks Anthropic AI model after Amazon flags security concern
Reports say Amazon’s warning about a Fable 5 safeguard bypass preceded export controls that forced Anthropic to shut the model for users.
By Maya Lindqvist · Senior Technology Correspondent
3 min read
The Trump administration used national-security export controls Friday to restrict Anthropic’s Fable 5 model and its underlying Mythos 5 system, Fortune reported. Fortune said the order forced Anthropic to disable the models for users, making it the first reported U.S. use of export controls to halt access to a commercial AI model already in public use.
Multiple media reports said Amazon CEO Andy Jassy warned senior administration officials Thursday after Amazon researchers found prompts that could make the Mythos-class model produce restricted cyberattack information. Reuters reported that Amazon raised concerns about Anthropic’s AI models before the government action.
Politico reported, citing an unnamed person familiar with Amazon’s talks, that the government had asked Amazon for feedback on Anthropic’s new model. Fortune said it remains unclear whether Amazon tested the model at the White House’s request or on its own.
An Amazon spokesperson previously told Fortune that governments often seek the company’s advice on possible security risks because Amazon is a major cloud provider serving public and private customers. The spokesperson said Amazon does not discuss details of those conversations.
Semafor reported, citing unnamed sources, that the U.S. government suspected a Chinese-linked group had already used the jailbreak that Amazon identified. Semafor also said it was unclear what evidence supported that suspicion, and an Anthropic spokesperson told Semafor that the White House did not raise Chinese access to Mythos in its discussions with the company. The spokesperson also said Anthropic bars access to its products from inside China.
Politico reported that Anthropic CEO Dario Amodei held several calls with senior administration officials after Amazon’s warning. According to Politico, Amodei argued that Amazon had found a narrow bypass rather than a full jailbreak of the model’s safeguards.
Fortune reported, citing a person familiar with Anthropic, that the company was given 90 minutes to withdraw its newest model and had not received earlier notice of a national-security threat. The Commerce Department order barred distribution of Fable 5 and Mythos 5 to foreign nationals, a category Fortune said includes people outside the U.S. and non-citizens working in the country, including Anthropic employees.
White House AI adviser David Sacks gave a different account on X over the weekend. Sacks said a trusted partner of Anthropic and the government had found a jailbreak in Fable 5’s guardrails, and that the administration asked Amodei to fix the issue or pull the model. Sacks said Amodei refused, which led the administration to issue the controls, and said the White House wants the model returned to general release once Anthropic fixes the issue.
Axios reported that an administration official said other models now on the market do not pose the same national-security risk because they have not reached Mythos 5’s capability level. The official told Axios that future models crossing that level would need government review before release.
The dispute also follows earlier friction between Anthropic and the Trump administration. Fortune reported that the administration had designated Anthropic a supply-chain risk for Pentagon contractors after the company declined contract terms allowing its models to be used for all lawful purposes. Fortune said Anthropic cited concerns about autonomous weapons and mass domestic surveillance and is challenging that designation in court.
The shutdown drew criticism from AI policy figures and politicians abroad. Dean Ball, an AI policy expert who briefly served in the Trump administration, criticized the move on X, while Ben Murphy of the Institute for Progress warned that it could discourage AI labs from being open with the government about model risks.
European and British politicians cited the decision as evidence of dependence on U.S. AI providers, Fortune reported. Former French prime minister Édouard Philippe compared AI to critical infrastructure, while U.K. lawmakers Al Carns and Tom Tugendhat said the cutoff showed the strategic risks of relying on foreign-controlled AI systems.
This story draws on original reporting from Fortune.