The US government has suddenly pulled the plug on Anthropic’s Fable 5 and Mythos 5 models, claiming national security concerns. All foreign nationals, including Anthropic employees abroad, are barred from accessing these AI models. The directive came with little explanation, although there’s chatter about a potential method to “jailbreak” Fable 5, revealing minor vulnerabilities that are quite common across AI models. Despite rigorous testing and collaboration with experts to secure Fable 5, the government fears remain. Anthropic argues that none of the identified vulnerabilities pose a serious threat and believes the government’s drastic action could cripple the industry’s progress if applied universally. Anthropic is complying, but they’re not thrilled—calling for a more transparent and fact-based approach to such shutdowns. They’re scrambling to restore access, asserting that the decision was unnecessary overreach.
Read more…
