Amazon CEO Andy Jassy might've been the one who blew the whistle on Anthropic's AI models — and the fallout was swift.
The Wall Street Journal reports that Jassy told Treasury Secretary Scott Bessent and other government officials that Amazon researchers used Anthropic's Claude Fable 5 to obtain information that could be used in cyberattacks. The government responded by imposing an export control ban on both the Fable 5 and Mythos 5 models on Friday.
Anthropic cut off worldwide access to the two models the same day.
Jassy's conversation with Bessent reportedly happened before the government crackdown. It's not clear exactly when they spoke, but the timing suggests Jassy's concerns directly triggered the Treasury's action.
The export ban means companies outside the US can't legally obtain or use Fable 5 and Mythos 5. That's a huge deal for Anthropic, which had been selling access to those models globally.
Anthropic is one of the most valuable AI startups in the world. It was founded by former OpenAI employees who wanted to build safer AI. The company has raised billions from investors including Google and — awkwardly — Amazon itself. Amazon has invested nearly $4 billion in Anthropic, making this a very uncomfortable situation for both companies.
Claude Fable 5 was Anthropic's most advanced model, released just months ago. It was marketed as being better at creative tasks and complex reasoning. Mythos 5 was a smaller, faster model aimed at developers.
The Wall Street Journal didn't specify exactly what information Amazon researchers extracted. But the phrase "could be used in cyberattacks" covers a lot of ground — from exploit code to vulnerability discovery methods to social engineering scripts.
This isn't the first time AI companies have faced questions about safety. But it's rare for a major investor like Amazon to go directly to the government with concerns about a portfolio company's product.
The export ban is also notable because it targets models that were already in wide use. Usually, export controls on AI target unreleased or future models. Banning existing products is an escalation.
Anthropic hasn't commented on Jassy's reported role. The company's public statement on Friday only said it was "cooperating with government officials" and that safety was its top priority.
Amazon declined to comment to the Journal. The Treasury Department also didn't respond to requests for comment.
What happens next is unclear. The export ban is in effect, but Anthropic could appeal or negotiate with the government. The company could also try to release modified versions of the models that don't trigger security concerns.
For now, developers and businesses that relied on Fable 5 and Mythos 5 are stuck. They can't access the models anymore, and there's no timeline for when — or if — they'll come back.
This story is developing.