Amazon CEO's Warnings Triggered Ban on Anthropic AI Models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Amazon CEO Andy Jassy told Trump administration officials including Treasury Secretary Scott Bessent that researchers had used prompts to get Anthropic's Fable 5 model to provide information that could aid cyberattacks and was supposed to be off-limits
- The Trump administration responded by halting all foreign use of Anthropic's most capable AI models; White House officials held a meeting and security researchers tested Amazon's claims before deciding the most direct fix was preventing foreign governments, companies, and individuals from accessing the tool
- President Trump signed off on the action, which aligned with long-standing administration doubts that Anthropic could be trusted to manage the security risks of its new model—a view reinforced by a Friday call with Anthropic CEO Dario Amodei
- Anthropic said it was shutting off access to its Mythos and Fable models to all users to ensure compliance, a move the company said could hamper companies worldwide that use the tools to identify software vulnerabilities
- The foreign-access rule effectively prevents many of Anthropic's foreign-born researchers from working on its latest models, according to the company
- Amazon is a major Anthropic investor that supplies the AI firm with data-center chips and deploys its best models to find software bugs, underscoring how tightly the two companies' fortunes are now bound by the new restrictions
Why it matters: The crackdown shows how a single CEO-to-administration conversation can rapidly translate into a sweeping national-security export restriction—Anthropic, which employs many foreign researchers and serves global clients, now faces restricted talent access and a narrower customer base for its most advanced models, while its largest backer (Amazon) effectively helped trigger the rule that constrains it.
