A new AI model from privately held Anthropic is so powerful that the company won’t release it publicly and the White House is reversing its hands-off approach to AI regulation, the Wall Street Journal reported.

Vice President JD Vance, who told a Paris AI summit last year that overregulation could kill the industry, warned the heads of Microsoft Corp (NASDAQ:MSFT), Alphabet Inc (NASDAQ:GOOGL), OpenAI and Anthropic on a recent call that Mythos, which can find software vulnerabilities on its own, could trigger cyberattacks on small-town banks, hospitals and water plants.

From Hands-Off To FDA For AI

President Donald Trump is weighing an executive order to create a formal oversight process for the most advanced AI models, a reversal from his December 2025 order targeting state AI laws.

National Economic Council Director Kevin Hassett likened the expected regime to FDA drug approval on Fox Business this week. National Cyber Director Sean Cairncross is leading the response, and Treasury Secretary Scott Bessent has warned banking executives directly about the model.

Polymarket’s “Trump orders federal review of AI model releases by May 31?” contract sits at 19%, suggesting traders aren’t yet pricing in an imminent order.

White House adviser David Sacks has publicly pushed back on the regulation push, arguing on his All-In podcast that the threat is overblown.

Anthropic Pulls Away From OpenAI

Anthropic and OpenAI are in a race to IPO that has gotten increasingly one-sided. While Altman’s problems mount: missed user targets, a $180 billion Musk lawsuit, and a CFO worried about paying for compute. Anthropic just developed a model the White House thinks is too dangerous to release.

Polymarket traders give Anthropic a 63% probability of having the top AI model at the end of June, with $5m in volume on the contract.

Anthropic is also a 68% favorite to IPO before OpenAI on a separate Polymarket market.

The Equity Read

Microsoft, OpenAI’s largest backer and the most direct public exposure to any OpenAI slowdown, is trading at $415. Amazon.com Inc (NASDAQ:AMZN), with $8 billion in Anthropic, is trading at $272.

Cybersecurity stocks are already moving on the AI threat narrative. Palo Alto Networks Inc (NASDAQ:PANW) and CrowdStrike Holdings Inc (NASDAQ:CRWD) jumped 5.4% and 7% on Thursday. KeyBanc lifted its CrowdStrike price target to $525 last month, citing a CIO survey that found cyber budgets benefiting directly from Mythos.

The shift lands days before Trump’s summit with Chinese leader Xi Jinping, where AI risk is on the agenda. Coordinated US-China restrictions could bake the regulation thesis in further.

Image: Shutterstock