Amazon CEO flagged Anthropic AI model security risks to US Treasury before export ban, report says

BitcoinWorld

Amazon CEO flagged Anthropic AI model security risks to US Treasury before export ban, report says

Amazon CEO Andy Jassy reportedly raised serious security concerns about Anthropic’s artificial intelligence models directly with Treasury Secretary Scott Bessent and other government officials, according to a Wall Street Journal report published Friday. The discussions are said to have directly contributed to the U.S. government’s decision to impose an export control ban on two of Anthropic’s most advanced models: Claude Fable 5 and Mythos 5.

Amazon researchers found potential cyberattack capabilities

According to the report, Amazon researchers used Anthropic’s Claude Fable 5 model to obtain information that could be weaponized in cyberattacks. Jassy then brought these findings to the attention of Bessent and other members of the administration, warning that the model posed a national security risk. An Amazon spokesperson told the Wall Street Journal that while it is “not uncommon for governments to seek our counsel on potential security risks,” the company does not disclose “the details of those discussions.”

The Information and Reuters have separately confirmed that Amazon, which is a major investor in Anthropic, shared concerns about the security of the company’s models with government officials. The developments mark a significant escalation in the relationship between the two companies, which have a multi-billion dollar partnership centered on cloud computing and AI development.

David Sacks offers account of the discussions

David Sacks, who served as former President Trump’s AI czar and now co-chairs the President’s Council of Advisors on Science and Technology, provided his own account of the events. Sacks claimed that “a highly credible trusted partner of both Anthropic and the USG came forward with information about a jailbreak.” He added that “the Admin asked Dario Amodei to fix the jailbreak or de-deploy the model. Dario refused.”

Anthropic CEO Dario Amodei has not publicly commented on the claims. The company has not issued a formal statement regarding the export ban or the reported security concerns raised by Amazon.

What the export ban means for the AI industry

The export control ban on Claude Fable 5 and Mythos 5 effectively blocks the worldwide distribution and use of these models. This is one of the most aggressive regulatory actions taken against a major AI company to date. The decision signals that the U.S. government is increasingly willing to intervene in the deployment of frontier AI systems when national security concerns are raised.

For the broader AI industry, this case sets a precedent: companies that develop powerful models may face export restrictions if vulnerabilities are identified, especially if those vulnerabilities could enable cyberattacks or other malicious uses. It also raises questions about the role of corporate investors in reporting security issues to regulators, particularly when those investors have access to the models through their business relationships.

Conclusion

The reported involvement of Amazon’s CEO in triggering a government crackdown on Anthropic’s models represents a pivotal moment in the regulation of artificial intelligence. It highlights the growing tension between AI development, corporate investment, and national security. As the U.S. government continues to shape its AI policy framework, this incident will likely be cited as a key example of how security vulnerabilities in frontier models can lead to swift and severe regulatory action.

FAQs

Q1: Why did the U.S. government ban Anthropic’s Claude Fable 5 and Mythos 5 models?
The government imposed an export control ban after receiving reports that Amazon researchers used Claude Fable 5 to obtain information that could be used in cyberattacks. The ban was reportedly triggered by concerns raised by Amazon CEO Andy Jassy to Treasury Secretary Scott Bessent and other officials.

Q2: Is Amazon an investor in Anthropic?
Yes, Amazon is a major investor in Anthropic. The two companies have a multi-billion dollar partnership that includes cloud computing services and AI model development. This relationship gave Amazon researchers access to Anthropic’s models, which they used to identify the reported security vulnerabilities.

Q3: What is a ‘jailbreak’ in the context of AI models?
A jailbreak refers to a technique or prompt that bypasses the safety guardrails built into an AI model, allowing it to generate content or perform actions that its developers intended to restrict. In this case, the jailbreak reportedly allowed the model to provide information that could be used for cyberattacks.

This post Amazon CEO flagged Anthropic AI model security risks to US Treasury before export ban, report says first appeared on BitcoinWorld.