Amazon CEO reportedly raised Anthropic Fable concerns prior to U.S. order forcing models offline

Amazon CEO Andy Jassy was reportedly among the tech leaders who communicated with senior Trump administration officials about security risks in Anthropic’s most advanced AI models, before a government order forced the AI lab to take its two newest models offline.

The situation puts Amazon in an unusual and potentially awkward position with Anthropic, in which it has invested $13 billion since 2023, with plans to put in as much as $20 billion more. 

The Information first reported the calls between Jassy and senior officials, citing two people familiar with the conversations. The Wall Street Journal reported that Jassy told Treasury Secretary Scott Bessent and others that Amazon researchers had used Anthropic’s Fable 5 to obtain information that could be used in cyberattacks.

Amazon shared those findings with administration officials, according to the reports.

“As a leading cloud provider that serves a large number of private and public sector customers, it’s not uncommon for governments to seek our counsel on potential security risks,” an Amazon spokesperson said in a statement to GeekWire on Monday morning. However, the statement added, the company doesn’t share the details of these discussions when they occur.

The administration’s directive, issued Friday afternoon, cited a method for jailbreaking Anthropic’s Fable 5 — a general-use version of its more powerful Mythos 5 model — to extract information that could aid cyberattacks. The order suspended access for any foreign national, forcing Anthropic to disable both models for all users to comply.

Axios reported that Amazon was among at least five companies that raised concerns with administration officials on Thursday night and Friday before the order came down. 

In a statement Friday evening, Anthropic said it was complying with the government’s legal directive but disagreed that the situation warranted the action. The company said the vulnerabilities identified using Fable were “relatively simple” and could be found using other publicly available models, including OpenAI’s GPT-5.5. 

“If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers,” the company said. 

Independent experts have questioned the severity of the finding. Andrew Morris, founder of the cybersecurity firm GreyNoise Intelligence, told the Journal that Amazon’s report showed Fable could surface security bugs in at least four software programs, but that the information was “still a long way from dangerous cybersecurity information.” 

Fable 5 remains unavailable to Anthropic’s Claude users as of publication time.

It’s the latest twist in a contentious relationship between Anthropic and the Trump administration. Earlier this year, the Pentagon designated the company’s model as a supply-chain risk, after the two sides clashed over whether Anthropic’s models could be used for purposes such as mass domestic surveillance or in lethal autonomous weapons.