Dario Amodei walks into the West Wing on Friday with something no amount of White House charm can fix: a government that banned his company and cannot stop needing it.
Anthropic's CEO is scheduled to meet White House chief of staff Susie Wiles on Friday, Reuters reported, confirming an Axios scoop — a breakthrough in the AI startup's eight-week fight with the Pentagon. But the meeting is not the real story. The real story is why it had to happen at all.
Three federal agencies are already working around the Pentagon's blacklist to get access to Anthropic's most powerful model, Mythos. Treasury and State have formally requested briefings, per Reuters. The Office of Management and Budget is drafting agency-by-agency protections to make the access legally defensible, The Next Web reported. The government spent two months trying to destroy Anthropic for refusing to remove safety guardrails. Its own agencies are now clearing the obstacles that same government put in place.
The collision is not new — Reuters first reported the agency courtship on April 16, and type0 has covered the Anthropic-Pentagon dispute extensively. What changes today is the level. This has escalated from bureaucratic workaround to CEO-level negotiation. The question is no longer whether agencies want Mythos. It is whether the White House will formally override the Pentagon's designation, or whether this remains a collection of individual agencies quietly pursuing a policy that officially does not exist.
Anthropic announced Mythos on April 7 as part of Project Glasswing, a controlled release program for what the company described as an emergent cybersecurity capability — not a feature they built, but something that appeared when reasoning and autonomy improved enough to find vulnerabilities at scale. The company said Mythos identified thousands of zero-day flaws across every major operating system and browser, a claim that drew skepticism from security researchers who noted the findings were based on 198 manual reviews. But the capability was serious enough that Anthropic did not release it publicly.
Project Glasswing now includes roughly 50 organizations: JPMorgan Chase, Amazon, Apple, Google, Microsoft, Nvidia, Cisco, CrowdStrike, and Palo Alto Networks, per The Next Web. Anthropic committed $100 million in usage credits and $4 million in direct donations to open-source security organizations. UK regulators at the Bank of England and the National Cyber Security Centre are separately examining what the model surfaced.
The Pentagon dispute began in February when Defense Secretary Pete Hegseth gave Amodei a Friday deadline to drop two safety restrictions: no use in fully autonomous weapons, and no deployment for mass surveillance of Americans. Amodei refused. Hegseth declared Anthropic a supply chain risk. President Trump ordered federal agencies to stop using the company's technology. A Pentagon official told Reuters Amodei had a "God complex." Trump called Anthropic a "radical left, woke company."
The courts produced a split decision. A federal judge in California issued a preliminary injunction blocking the designation, writing that "nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the US for expressing disagreement with the government," per The Next Web. A federal appeals court in Washington, D.C. denied Anthropic's request to temporarily halt the blacklisting while the case proceeds, Quartz reported. The practical result: Anthropic is barred from Pentagon contracts but can work with civilian agencies while the California injunction holds.
It is into that gap that Treasury Secretary Scott Bessent and Fed Chair Jerome Powell stepped. They urged executives at JPMorgan, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley to test Mythos for cybersecurity vulnerabilities, The Next Web reported — a recommendation that sits in direct contradiction to the administration's own policy.
The question the Wiles-Amodei meeting has to answer is not whether the White House likes Anthropic. It is whether the government can sustain a policy it designed to punish the company when three of its own agencies have concluded they cannot operate without what Anthropic built.
If Wiles and Amodei reach a binding agreement, it effectively supersedes the Pentagon's blacklist and sets a precedent that ethics commitments can be negotiated under pressure. If they do not, agencies continue operating in a legal gray zone — using a tool the government officially banned, through protections that a different branch of the same government is still drafting.
The meeting is scheduled for Friday. The policy incoherence has been running for eight weeks.
