The Security Alert That Triggered the Ban
Jassy reportedly informed Treasury Secretary Scott Bessent and other government officials about potential security risks tied to Anthropic’s AI models. According to The Wall Street Journal, Amazon researchers allegedly used Anthropic’s Claude Fable 5 model to obtain information that could be leveraged in cyberattacks. This communication from Jassy appears to have directly influenced what happened next.
Following Jassy’s report, the U.S. government imposed export controls targeting Anthropic’s Claude Fable 5 and Mythos 5 models. The move led Anthropic to cut off worldwide access to both models on Friday. An Amazon spokesperson confirmed that governments frequently seek the company’s counsel on potential security risks, though Amazon does not share details of those discussions publicly.
The core concern stemmed from Amazon researchers successfully exploiting Claude Fable 5 for potentially malicious purposes, pointing to a jailbreak capability that allowed the model to bypass its intended safety protocols. The Information and Reuters both reported separately that Amazon had conveyed similar concerns about the security of Anthropic’s models to government officials.
Two Very Different Stories
Not everyone agrees on how this unfolded. David Sacks, former Trump AI czar and co-chair of the President’s Council of Advisors on Science and Technology, offered a different account. He claimed that a highly credible trusted partner
presented a jailbreak vulnerability to both Anthropic and the U.S. government, and alleged that Anthropic CEO Dario Amodei refused to either fix the issue or pull the model when the administration requested it.
Anthropic tells a different story. In a blog post, the company stated that the capabilities causing government concern are already available in other publicly accessible models, suggesting the singling out of its models may not reflect a unique risk.
Ripple Effects Beyond Anthropic
The incident underscores the growing tension between rapid AI innovation and the need for robust security measures. AWS, Amazon’s own cloud computing division, has also been affected by the model cutoff, showing the operational impact extends well beyond Anthropic itself.
What This Means Going Forward
The government’s swift move to impose export controls signals an increasingly proactive stance on AI safety regulation. This event sets a notable precedent for how security vulnerabilities in advanced AI models might be handled at a national level going forward. Other AI developers and cloud providers may now find themselves reassessing how they engage with government bodies on AI security, as the relationship between leading tech companies and regulators enters a more critical phase.
Follow Hashlytics on Bluesky, LinkedIn , Telegram and X to Get Instant Updates
