White House forces Anthropic to disable new frontier models following abrupt export ban

The U.S. Commerce Department issued an export control order banning foreign nationals—including allies and Anthropic’s own international employees—from accessing its latest AI models, Fable 5 and Mythos 5. In response, Anthropic disabled global access to comply with the restrictions, which were triggered by national security concerns over potential jailbreak vulnerabilities. The decision came after reports from Amazon and the U.K.’s AI Safety Institute (AISI) highlighted risks in the models. Amazon’s CEO, Andy Jassy, alerted U.S. officials that researchers had extracted information from Fable 5 that could aid cyberattacks. Meanwhile, the AISI’s Red Team lead, Xander Davies, posted on X that they developed a jailbreak within hours of accessing the model, allowing malicious responses and multi-step tool-calls. Anthropic claimed the risks were overstated, arguing the flaw was minor and existed in rival models like OpenAI’s GPT-5.5. The company also warned that the U.S. government’s approach could stifle innovation in frontier AI. However, the White House had previously requested Anthropic’s CEO, Dario Amodei, to patch or de-deploy Fable 5, which he reportedly declined. The ban has sparked criticism from allies, particularly in Europe. The European Commission is assessing the consequences, with spokesperson Thomas Regnier noting concerns about the decision’s broader impact. Meanwhile, Anthropic faces reputational damage and operational setbacks, as its own employees and enterprise customers lose access to the models. The incident underscores growing tensions between AI innovation and regulatory oversight, raising questions about how governments will balance security risks with technological progress.