Amazon chief executive Andy Jassy is reportedly at the centre of security concerns that led Anthropic to withdraw global access to two of its artificial intelligence models.
According to reports, Jassy informed US Treasury Secretary Scott Bessent and other government officials that Amazon researchers had used Anthropic’s Claude Fable 5 model to obtain information that could potentially be exploited in cyberattacks. The disclosure is said to have contributed to the US government imposing export-control restrictions on both the Fable 5 and Mythos 5 models.
In a statement, an Amazon spokesperson said it is “not uncommon for governments to seek our counsel on potential security risks”, but declined to comment on the substance of any discussions. The company also noted that Amazon Web Services (AWS) has itself been affected by Anthropic’s decision to restrict access to the models.
Reports from The Information and Reuters similarly indicated that Amazon, one of Anthropic’s largest investors, had raised concerns with authorities regarding the security implications of the company’s AI systems.
David Sacks, who previously served as the Trump administration’s AI czar and now co-chairs the President’s Council of Advisors on Science and Technology, offered further insight into the matter. He claimed that a “highly credible trusted partner” of both Anthropic and the US government had identified a jailbreak capable of bypassing the model’s safeguards.
According to Sacks, government officials asked Anthropic chief executive Dario Amodei to address the vulnerability or withdraw the model from deployment, but he allegedly declined to do so.
Anthropic, however, has pushed back against the concerns, arguing in a blog post that the capabilities cited by officials are already available in several other publicly accessible AI models.
