Amazon CEO Andy Jassy Raised Security Concerns Before US Shut Down Anthropic's Top AI Models

Amazon's Quiet Lobbying Preceded an Unprecedented Shutdown

Amazon CEO Andy Jassy held discussions this week with Treasury Secretary Scott Bessent and other senior Trump administration officials about security vulnerabilities in Anthropic's most advanced AI models. Those conversations directly preceded the Commerce Department's export-control directive issued on Friday, 12 June, ordering Anthropic to suspend global access to its Claude Fable 5 and Mythos 5 systems. The revelation, first reported by The Information and confirmed by the Wall Street Journal, exposes a striking conflict of interest: Amazon is simultaneously Anthropic's largest investor, with over $4 billion committed, its primary cloud infrastructure partner through AWS, and now the company whose lobbying helped trigger a government shutdown of Anthropic's flagship products.

The Directive and Its Fallout

Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei on Friday placing both models under export controls, barring access for all foreign nationals inside and outside the United States, including Anthropic's own foreign-national employees. Anthropic said it received the directive at 5:21 p.m. ET and, unable to verify citizenship for all users at such short notice, disabled both models for every customer worldwide. "The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," the company stated, adding that access to its other models remains unaffected. An administration official told Axios the restrictions may last "the next few weeks" while security concerns are evaluated.

Competing Narratives

At the centre of the dispute is a claimed jailbreak of Claude 3.5, reportedly demonstrated by researchers who used the model to extract information useful for cyberattacks. Anthropic has characterised it as a "potential narrow, non-universal jailbreak" involving previously known vulnerabilities, arguing that the model was merely prompted to read a specific codebase and identify software flaws. David Sacks, former White House AI adviser, said Amodei refused a government request to "fix the jailbreak or de-deploy the model" after "a highly credible trusted partner" reported the vulnerability. Anthropic disputed the scope of the concern, calling the directive a "misunderstanding" and saying it hopes to restore access as soon as possible.

A Tangled Web of Interests

The episode highlights extraordinary entanglements in the AI industry. Amazon competes with Anthropic through its own Nova models while simultaneously serving as the company's cloud provider and largest backer. Cybersecurity expert Katie Moussouris, who reviewed the jailbreak report, called the government's response "a complete overreaction," while a Democratic House member termed the directive "attempted corporate murder." Fable 5 launched publicly on 9 June as a version of the more powerful Mythos 5, which Anthropic had restricted since April to vetted cybersecurity defenders due to its ability to identify zero-day vulnerabilities. The broader Mythos model had already drawn alarm from federal officials, prompting Bessent and Federal Reserve Chair Jerome Powell to warn bank CEOs about its risks in April.