Anthropic restricts release of Mythos model, citing national security risks Odaily Planet Daily News: Anthropic has decided to restrict the public release of the Mythos model due to its highly automated network attack capabilities. The report states that the model has been able to independently complete the vulnerability discovery and exploitation process in internal testing, and generate multi-step attack plans. According to informed sources, Mythos was able to independently build intrusion tools for Linux systems in early testing and complete complex vulnerability chain attacks under guidance. The relevant capabilities have been assessed as potentially posing risks to global infrastructure. Anthropic management ultimately positioned Mythos as a network defense tool and opened testing to some institutions through restricted means. Industry insiders point out that similar models may significantly improve network attack and defense efficiency, but may also bring new security challenges.