Anthropic has made a sudden and dramatic move, discontinuing access to its newly launched Mythos 5 and Fable 5 models, in response to a directive from the US Commerce Department. The decision comes with immediate effect, leaving users of these artificial intelligence tools in limbo.

The move is driven by concerns over potential security risks, particularly related to jailbreaks that could compromise safety measures designed to prevent misuse of the technology. For now, other Anthropic models remain accessible, but this development signals a shift towards more stringent oversight and control.

According to reports from Axios, the administration believes there is evidence of a narrow, non-universal jailbreak in Fable 5, which can review codebases for software flaws. While Anthropic maintains that such jailbreaks have only led to minor vulnerabilities, the broader implications for AI safety and regulation are significant.

The directive from the US Commerce Department suggests that gaining time is crucial as they aim to 'harden' their national security apparatus against these threats. This process could take just a few weeks, providing some relief but also highlighting the ongoing challenges in managing and regulating advanced AI technologies.