Dario Amodei, a leader at Anthropic, stated that their AI model, Mythos, was so powerful that it had to be secured for national security reasons. He noted that the model revealed a significant increase in vulnerabilities that surprised the team. This development comes amid the Trump administration’s directive to impose restrictions on foreign access to advanced AI models, including those developed by Anthropic, to mitigate cybersecurity risks.

Mythos: Mythos is Anthropic’s advanced frontier AI model focused on general capabilities with standout performance in cybersecurity tasks. The company restricted broader access to the model after internal testing revealed its potential for misuse in vulnerability discovery. In the reported comments, Dario Amodei described Mythos as requiring lockdown measures due to national security implications from its unexpected capabilities.
Dario Amodei: Dario Amodei serves as CEO and co-founder of Anthropic, where he oversees development of frontier AI systems. He has publicly addressed the security risks of powerful models in recent policy discussions. His statements in the news directly reference the decision to restrict Mythos access stemming from its performance in finding vulnerabilities.

AI Policy: Anthropic’s leadership has called for mandatory third-party testing of frontier AI models to address risks in cybersecurity and other domains.
National Security: The Trump administration has ordered restrictions on foreign access to certain advanced AI models developed by Anthropic due to cybersecurity and national security risks.