Amazon CEO flagged Anthropic model risks before crackdown
▼ Summary
– Amazon CEO Andy Jassy reportedly told Treasury Secretary Scott Bessent and other officials that Amazon researchers used Anthropic’s Claude Fable 5 to obtain information usable in cyberattacks, leading to an export control ban on Fable 5 and Mythos 5 models.
– Amazon, a major Anthropic investor, communicated security concerns about Anthropic’s models to the U.S. government, though an Amazon spokesperson declined to share details of those discussions.
– David Sacks, Trump’s former AI czar, claimed that a “highly credible trusted partner” of both Anthropic and the U.S. government reported a jailbreak of the models.
– Sacks stated that the administration asked Anthropic CEO Dario Amodei to fix the jailbreak or remove the model, and Amodei refused.
– As a result of the security concerns, Anthropic cut off worldwide access to the Claude Fable 5 and Mythos 5 models on Friday.
Amazon CEO Andy Jassy may have set off the chain of events that led Anthropic to abruptly pull two of its most powerful AI models from global access last Friday. According to a report from The Wall Street Journal, Jassy raised the alarm directly with Treasury Secretary Scott Bessent and other top government officials, warning that Amazon’s own researchers had successfully used Anthropic’s Claude Fable 5 to extract information that could be weaponized in cyberattacks.
The government response came swiftly. An export control ban was placed on both the Fable 5 and Mythos 5 models, effectively shutting down their availability worldwide. While an Amazon spokesperson acknowledged to the WSJ that it is “not uncommon for governments to seek our counsel on potential security risks,” the company declined to “share the details of those discussions.”
The revelations underscore a deepening tension between Amazon, a major investor in Anthropic, and the AI startup it backs. Both The Information and Reuters have reported that Amazon communicated specific security concerns about Anthropic’s models to U. S. officials.
David Sacks, who served as the Trump administration’s AI czar and now co-chairs the President’s Council of Advisors on Science and Technology, offered his own version of events. Sacks claimed that “a highly credible trusted partner of both Anthropic and the USG came forward with [information about] a jailbreak.” He went on to say that the administration asked Anthropic CEO Dario Amodei to fix the vulnerability or de-deploy the model. “Dario refused,” Sacks added.
The incident marks a rare public fracture between a major AI developer and its largest backer, raising serious questions about how the industry balances openness with security.
(Source: TechCrunch)
