U.S. officials press top AI CEOs on model safety as Anthropic rolls out “Mythos”

High-level call as new model debuts

It has been reported that U.S. Vice President and Treasury Secretary Janet Yellen? — wait: the original article names "Vice President Vance and Treasury Secretary Bessent" (these appear to be placeholders or mis-transliterations) — convened a phone call last week with several leading AI and cybersecurity CEOs to question them on the safety and misuse risks of large language models. CNBC reported the call, and Chinese tech site 凤凰网科技 (ifeng) carried the same account. Who was on the line? It reportedly included Anthropic CEO Dario Amodei, xAI CEO Elon Musk, Google CEO Sundar Pichai, OpenAI CEO Sam Altman, Microsoft CEO Satya Nadella, CrowdStrike CEO George Kurtz and Palo Alto Networks CEO Nikesh Arora.

Focus on cyber risk and deployment controls

Officials pressed CEOs on the safety posture of their large language models and on contingency plans if models evolve in ways that favor attackers — for example enabling cyber intrusions. It has been reported that Anthropic briefed senior U.S. officials on the full capabilities of its new Mythos model, including possible applications to both cyber attack and defense, and said it was willing to support government testing and evaluation. Anthropic also reportedly launched Mythos this week to a limited set of enterprise partners that include Apple, Google, Microsoft, NVIDIA, Palo Alto Networks and CrowdStrike.

Political and regulatory backdrop

This engagement comes amid heightened scrutiny from the U.S. government. Treasury officials and Federal Reserve Chair Jerome Powell separately summoned major banks to discuss potential threats tied to advanced models, and the Trump administration has moved to limit federal use of Anthropic’s Claude platform while the company pursues legal challenges to a Department of Defense supply‑chain risk designation — rulings on those challenges have been inconsistent across U.S. courts. Reportedly, the episode underscores a broader question: how will policymakers balance innovation with national‑security and cyber‑risk concerns as more powerful AI systems reach the market?