Amazon CEO Andy Jassy appears to have been the source of security concerns that prompted the U.S. government to impose export controls on two Anthropic AI models last week, according to reporting from The Wall Street Journal. Jassy reportedly informed Treasury Secretary Scott Bessent and other federal officials that Amazon researchers had used Anthropic's Claude Fable 5 model to extract information potentially useful for cyberattacks. Following those communications, the government placed an export ban on the Fable 5 and Mythos 5 models, and Anthropic responded by cutting off worldwide access to both on Friday. The Information and Reuters corroborated the account, noting that Amazon is a major investor in Anthropic.
An Amazon spokesperson declined to confirm the specifics but said in a statement that "it's not uncommon for governments to seek our counsel on potential security risks," while emphasizing that the company does not "share the details of those discussions." The spokesperson also pointed to an update indicating that Amazon Web Services has been impacted by the model cutoff. David Sacks, who served as President Donald Trump's AI czar and now co-chairs the President's Council of Advisors on Science and Technology, offered his own version of events on social media, claiming that "a highly credible trusted partner of both Anthropic and the USG came forward with a jailbreak." Sacks added that the administration then asked Anthropic CEO Dario Amodei to patch the vulnerability or pull the model, and that "Dario refused."
Anthropic pushed back on the government's characterization in a blog post, arguing that the capabilities allegedly raising red flags are already accessible through other publicly available models. The company has not publicly detailed the specific jailbreak or security issue at the center of the dispute, but the standoff underscores growing tensions between the leading AI lab and federal regulators over how powerful models should be controlled before reaching global users. The export restrictions mark one of the most aggressive actions the U.S. government has taken to constrain the international spread of frontier AI systems, and they raise questions about whether similar constraints could be applied to other developers whose models demonstrate comparable capabilities.