ANTHROPIC BLOCKS FABLE, MYTHOS AFTER US ORDER

Anthropic complied with the directive Friday evening, citing national security concerns raised by the Trump administration. The order blocks access for foreign nationals both inside and outside the US, including Anthropic employees. The government's decision followed cybersecurity research from Amazon that identified a potential vulnerability in Fable 5. According to the Wall Street Journal, Amazon researchers discovered they could manipulate the model through specific prompts to extract information potentially useful for cyberattacks. CEO Andy Jassy reportedly discussed the findings with White House officials, triggering the export control directive. AnthropicDisputed the government's characterization of the threat. In a statement, the company said it was complying while noting that authorities "did not provide specific details of its national security concern." The firm argued the jailbreak method was narrow in scope and that similar capabilities exist in other widely available AI systems. The order marks a significant moment for AI regulation. It sets potential precedent for how the US government will restrict advanced model access, with implications extending to competitors like OpenAI, Google, and Meta. The shutdown affects all customers currently using the models. Anthropic has not announced a timeline for restoring access or modifications that might allow limited foreign access in the future. This development reflects growing tension between AI development and national security policy. While the government cited export control concerns, questions remain about whether the identified vulnerability warrants a complete model suspension versus targeted restrictions. The incident underscores how AI capabilities are increasingly subject to regulatory scrutiny at the highest government levels.