Anthropic Fable 5: Amazon CEO Flags AI Security Flaws

Amazon's CEO, Andy Jassy, alerted the White House to security flaws in Anthropic's Fable 5 AI model. This led to an export ban, forcing Anthropic to disable both Fable 5 and Mythos 5 globally. Here's what happened: Amazon researchers used prompts to make Fable 5 provide information useful for cyberattacks. Jassy reported these findings to Treasury Secretary Scott Bessent. On June 12, the Commerce Department issued a directive, banning all foreign nationals from accessing Fable 5 and Mythos 5. Anthropic called this response disproportionate, stating the jailbreak revealed only minor vulnerabilities also present in other models. However, former AI czar David Sacks said officials asked Anthropic to fix the issue or remove the model, but Anthropic's CEO, Dario Amodei, refused. This situation is notable because Amazon is Anthropic's largest investor, with a $13 billion stake. The bottom line is that concerns about AI model security can lead to immediate and significant government intervention.