White House Called Emergency Meeting with Tech CEOs Before Anthropic Mythos Release Over AI Security Fears
The Trump White House convened an emergency call with the CEOs of America's leading AI and cybersecurity companies in the week before Anthropic released its Mythos model — questioning them directly about AI model security and the risk that advanced AI could scale cyber attack capabilities in favour of adversaries. The private call, led by Vice President JD Vance and Treasury Secretary Scott Bessent, reveals that the federal government was tracking Mythos as a potential national security event before it was publicly announced. It reflects the same concerns that have driven the National Cyber Director's parallel push to identify AI-exploitable vulnerabilities in critical infrastructure.
Who Was on the Call and What Was Discussed
Participants included Dario Amodei (Anthropic), Sam Altman (OpenAI), Sundar Pichai (Google), Satya Nadella (Microsoft), Elon Musk (xAI), George Kurtz (CrowdStrike), and Nikesh Arora (Palo Alto Networks). The discussion focused on AI model security posture, safe deployment practices, and how to respond if AI systems "scale in favour of attackers" — a specific concern about Mythos given its documented offensive and defensive cybersecurity capabilities.
Anthropic confirmed it had proactively briefed senior U.S. government officials on Mythos Preview's full capabilities — including its ability to reason about offensive cyber operations — before any external release. According to reporting by CNBC, the company's position was explicit: "Bringing government into the loop early — on what the model can do, where the risks are, and how we're managing them — was a priority from the start." Mythos launched to a limited group of partners including Apple, Google, Microsoft, Nvidia, Palo Alto Networks, and CrowdStrike before broader availability.
Banks Were Also Briefed — Separately
In a parallel development, Treasury Secretary Bessent and Federal Reserve Chair Jerome Powell called a surprise meeting with heads of major U.S. banks to address the threat that Mythos poses to financial sector cybersecurity. The involvement of the Fed — not a traditional actor in AI governance — signals that U.S. regulators view Mythos's capabilities as a systemic financial risk, not just a technology policy question. This connects to the parallel track on which Goldman Sachs and Citigroup are already testing Mythos internally for financial applications — creating an unusual situation where banks are simultaneously potential users and potential targets of the model's capabilities.
Anthropic's Complicated Federal Relationship
The White House engagement comes against a complicated legal backdrop. Anthropic lost a bid to reverse its Department of Defense blacklisting in a federal appeals court — meaning it cannot pursue DOD contracts despite being cleared to work with other federal agencies. The split rulings create an awkward dynamic: Anthropic is briefing the White House, the Fed, and major banks on its most powerful model while simultaneously being locked out of direct defense contracts. Anthropic's recent multi-gigawatt compute deal with Google and Broadcom underscores how rapidly the company is scaling capabilities that the government is now scrambling to understand.
Frequently Asked Questions
Why did the White House call tech CEOs before Anthropic released Mythos?
The White House, led by VP JD Vance and Treasury Secretary Bessent, convened the call to assess AI model security risks — specifically whether Mythos's cyber capabilities could be exploited by adversaries or used to scale attack capabilities beyond current defenses. The call included Anthropic, OpenAI, Google, Microsoft, xAI, CrowdStrike, and Palo Alto Networks.
What are Mythos's cybersecurity capabilities that concerned the government?
Mythos has documented capabilities in both offensive and defensive cybersecurity reasoning — it can analyze and reason about attack vectors and defenses at a level that government officials feared could give bad actors a meaningful capability upgrade if the model were widely misused or leaked.
Can Anthropic work with the US government after its DOD blacklisting?
Anthropic remains blocked from Department of Defense contracts after losing an appeals court ruling. However, it can still work with other federal agencies — and has actively engaged with the White House, Treasury, and the Fed on Mythos's capabilities and safety measures.
The Bottom Line
The White House's emergency call before Mythos's release is the clearest signal yet that the U.S. government now treats frontier AI model launches as national security events. Anthropic's decision to proactively brief government officials — and the administration's parallel outreach to bank regulators — signals that the gap between AI development and AI governance is narrowing, even if it has not yet closed.