The U.S. government has reversed its two-week-old order blocking Anthropic’s most advanced AI systems, Claude Mythos 5 and Fable 5, now allowing ‘certain trusted’ companies to access them. The Trump administration’s shift follows mounting industry pressure and national security concerns over keeping cutting-edge AI from adversaries.
Anthropic expressed satisfaction with the decision, stating, ‘We are pleased.’ The reversal underscores the delicate balance the administration must strike between fostering American AI leadership and preventing sensitive technology from reaching rivals like China. The caveat—limiting access to vetted firms—reflects ongoing anxiety about AI’s dual-use potential.
The original ban had forced Anthropic to take the models offline globally. Now, only approved entities can use the systems, though the precise criteria for trustworthiness remain undisclosed. This selective reopening aims to maintain U.S. competitiveness without risking technology leaks.
Meanwhile, a thriving underground economy for Claude access persists in China. According to Wired, users employ ‘transfer station’ sites that buy API tokens abroad and distribute them locally, along with proxy services and fake identities sourced on Telegram. This parallel market highlights enforcement challenges.
Critics argue the partial reversal may not stem the tide of unauthorized access and could embolden workarounds. The administration’s next steps will likely involve stricter monitoring of approved users and renewed efforts to disrupt the Chinese shadow market.