Amazon CEO's Talks Reportedly Prompted U.S. Crackdown on Anthropic AI Models
Discussions between Amazon Chief Executive Andy Jassy and U.S. officials, including Treasury Secretary Scott Bessent, reportedly led the Trump administration to halt foreign use of Anthropic's most capable AI models. Amazon researchers allegedly discovered that Anthropic's Fable 5 model could provide information useful for cyberattacks. Following these reports, White House officials convened meetings, tested Amazon's claims, and directed Anthropic to address the identified vulnerabilities. The administration ultimately decided to prevent foreign governments, companies, and individuals from accessing the tool.
The Trump administration's decision to restrict all foreign access to Anthropic's most advanced AI models was reportedly initiated by conversations between Amazon Chief Executive Andy Jassy and U.S. officials, including Treasury Secretary Scott Bessent.
According to individuals familiar with the situation, Amazon researchers utilized a series of prompts to elicit information from Anthropic's Fable 5 model that could potentially aid cyberattacks. This information was intended to be off-limits, a detail Jassy conveyed to officials.
Following these reports, White House officials held a meeting to discuss a response, leading security researchers to test Amazon's claims. The administration subsequently requested Anthropic to either resolve the vulnerabilities or remove the model. Officials concluded that the most direct way to mitigate the risk was to prevent foreign entities from accessing the tool.
President Trump later approved the action, despite expressing reservations about its potential impact on innovation, a senior White House official stated. The administration had reportedly harbored long-standing concerns that Anthropic, a leader in U.S. AI development, could not adequately manage the security risks associated with its new model. A subsequent call between administration officials and Anthropic Chief Executive Dario Amodei reportedly reinforced these concerns.
Anthropic has publicly characterized the vulnerabilities flagged by Amazon as relatively basic. The company asserts that other publicly available models are capable of discovering similar issues and that these do not constitute a full "jailbreak." This perspective is shared by some security researchers familiar with Amazon's findings.
Amazon is a significant investor in Anthropic and supplies the company with chips for its data centers. (Source: Slashdot)
