The US government has issued an export ban on Anthropic's Mythos-class AI models. This regulatory action prevents non-US nationals from accessing these specific systems. The decision follows national security concerns raised by the technology involved.
Regulatory action follows jailbreak demonstration by Amazon team
Anthropic released two versions of the restricted software: Fable 5 and Claude Mythos 5. These models were forcibly taken offline after researchers at Amazon demonstrated they could be jailbroken. The breach allowed the generation of malicious code, triggering immediate intervention.
Amazon holds a $13 billion investment stake in Anthropic and played a key role in the response. CEO Andy Jassy contacted White House officials to report the security vulnerabilities. Treasury Secretary Scott Bessent was among the government figures notified about the risks.
The collaboration between Amazon and Anthropic includes a commitment for Anthropic to spend more than $100 billion on AWS technologies over ten years. This agreement covers current and future generations of Trainium silicon and Graviton CPU cores. The spending aims to provide superior price performance for AI workloads.
Source: TweakTown
