The Wall Street Journal has the inside story on what really triggered this week’s White House AI policy whiplash. The sudden about-face started with a high-level secret call initiated by Vice President Vance in April. Vance, along with the Treasury Secretary and Secretary of State, convened the CEOs of OpenAI, Anthropic, Google, Microsoft, and Elon Musk.
During the call, Vance warned that Anthropic’s previously undisclosed security model, Mythos, was too powerful — and if it got out of control, it could spark cyberattacks on local banks and water utilities that local governments simply aren't equipped to handle. That single conversation created chaos inside the White House and ultimately torpedoed its earlier pledge to step back from regulation.
The fallout has landed squarely on Anthropic. The White House explicitly told the company to pause expanding access to Mythos for critical infrastructure organizations, and formally tapped National Cyber Director Sean Cairncross to lead efforts to restrict the model. Meanwhile, the Treasury Department issued targeted risk warnings to banking executives. On the other side, OpenAI — which recently previewed a similarly capable model called GPT-5.5-Cyber — proactively consulted the government and implemented access restrictions, so it hasn't faced the same level of pushback.