Amazon raised security concerns about Anthropic models before crackdown

Amazon’s deep wager on Anthropic has collided with a rare public display of caution from inside Big Tech. Andy Jassy was among the technology leaders who raised security concerns with senior Trump administration officials about Anthropic’s most advanced models, even as Amazon remained one of the startup’s largest backers. The episode exposes a widening gap between private risk assessments and the race to deploy ever more powerful AI systems.

Anthropic said the U.S. government issued an export-control directive on June 12 ordering it to suspend access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States, including foreign-national employees. The company said it received the directive at 5:21 p.m. ET and disabled access to both models for all customers worldwide to comply. Anthropic also said the government appeared to believe someone had found a jailbreak method, but that the demonstrated issues were narrow, minor vulnerabilities that could also be found in other publicly available models.

The crackdown followed a frantic 24-hour scramble inside the White House. Politico reported that Anthropic chief executive Dario Amodei was in tense calls with Treasury Secretary Scott Bessent, White House Cyber Director Sean Cairncross and chief of staff Susie Wiles, while Jassy raised concerns to the White House on Thursday, two days after the model’s public release. By Friday morning, the matter had reached the highest levels of the White House, underscoring how quickly a technical worry about guardrails became a national-security decision.

The business backdrop makes the episode especially striking. In April, Amazon said it would invest an additional $5 billion in Anthropic, with the option to invest up to $20 billion more, after already putting in $8 billion. Anthropic, in turn, said it would spend more than $100 billion over 10 years on Amazon Web Services technologies, while more than 100,000 customers ran Anthropic Claude models on AWS. That interdependence means the same company financing Anthropic’s growth was also, according to people familiar with the discussions, feeding Washington’s concerns about the company’s newest models.

The order was unusual because it targeted access to AI software itself, not chips or other hardware, a sign that Washington’s export-control playbook is moving closer to deployed models. It also came just days after Anthropic launched Fable 5 on June 9, after earlier warnings about the hacking capabilities of Mythos 5 had already led the company to hold that model back from broad release. The result is a rare snapshot of the frontier AI industry under pressure from both sides at once: commercial partners pushing scale, and the government, with help from those same partners, testing where the safety line should be drawn.