Hegseth, White House allies intensify attacks on Anthropic

Defense Secretary Pete Hegseth criticized Anthropic after the AI company disabled access to its Fable 5 and Mythos 5 models to comply with a Trump administration export control directive, citing a 'jailbreaking' vulnerability. The company argued the directive was unnecessary and would disrupt the entire AI industry, while allies of the administration urged compliance with safety requests." "article": "The Trump administration intensified its scrutiny of Anthropic after the AI company disabled access to its Fable 5 and Mythos 5 models on Friday. The move followed a directive from the administration to suspend all access to the models by foreign nationals, which Anthropic said it had to enforce to comply. The company stated that the government discovered a method to bypass security measures in Fable 5, exposing minor vulnerabilities that were also present in other publicly available models. Defense Secretary Pete Hegseth reinforced the administration’s stance, declaring on X that the Pentagon had permanently barred Anthropic from its facilities after designating it a supply chain risk earlier this year. Hegseth’s criticism came after Anthropic’s CEO, Dario Amodei, had previously assured the government that its AI tools would not be used for fully autonomous lethal weapons or mass domestic surveillance. The latest clash stems from President Trump’s recent executive order requiring AI labs to submit models for government testing up to 30 days before public release. Amodei supported the order, proposing mandatory third-party audits to assess risks like cybersecurity threats, biological weapons, and loss of AI control. However, Anthropic argued that the new directive was overly broad and would halt all new model deployments across the industry. The company insisted that unsafe AI deployments should be blocked through a transparent, statutory process rather than unilateral government actions. David Sacks, a former Trump adviser on AI, criticized Anthropic’s response, stating that its refusal to comply contradicted its self-proclaimed focus on AI safety. Sacks suggested the administration expects Anthropic to address the security issue and lift the export control restrictions promptly. Democratic Sen. Mark Kelly also supported the directive, though his full statement was cut off in the source. The dispute highlights growing tensions between the AI industry and the Trump administration over regulation, security, and the pace of AI development.