Defense Secretary Pete Hegseth wanted to make an example of Anthropic. Two months later, the company's revenue reportedly passed OpenAI's.
In late February, the Pentagon threatened to designate Anthropic a "supply chain risk" and force military contractors to drop the company entirely. Anthropic was willing to work with the Defense Department on classified systems, but wouldn't sign off on mass surveillance of American citizens or autonomous weapons that can kill without a human in the loop.
In a February 27 statement, the company was blunt: "No amount of intimidation or punishment from the Department of War will change our position on mass domestic surveillance or fully autonomous weapons."
CEO Dario Amodei pointed out that the government's own threats didn't hold together. One arm of the Pentagon was calling Claude a security risk while another was calling it essential to national security. "These latter two threats are inherently contradictory," he wrote.
Steve Bannon, of all people, said Anthropic "had it right" in rejecting the deal. OpenAI, which signed its own agreement with the Defense Department, said it didn't understand why Anthropic "could not reach this deal." Outside analysts noted that OpenAI's agreement left key autonomous weapons concerns unresolved.
Users picked a side, too. ChatGPT uninstalls surged 295% the weekend after the news broke, while Claude hit No. 1 on the U.S. App Store.
Anthropic's revenue run rate was $14 billion in mid-February, already growing tenfold annually. By April, multiple industry trackers reported it had climbed to roughly $30 billion, reportedly surpassing OpenAI. The company closed a $30 billion Series G at a $380 billion valuation and has attracted investor interest at $800 billion. Axios called it growth with no precedent in American corporate history.
Anthropic also launched Project Glasswing, a cybersecurity initiative built around its unreleased Mythos model, assembling a coalition of more than 50 organizations including Apple, Google, Microsoft, Nvidia, CrowdStrike and Palo Alto Networks. Lee Klarich, Palo Alto Networks' Chief Product and Technology Officer, said his team used Mythos to "identify complex vulnerabilities that prior-generation models missed entirely." Anthropic's own Frontier Red Team says the model's capabilities aren't close to plateauing.
By mid-April, the White House appeared ready to come to the table. Treasury Secretary Scott Bessent and Chief of Staff Susie Wiles met with Amodei, and the administration is now reportedly considering giving federal agencies access to Claude Mythos.
Anthropic's defiance gave it something no marketing budget could buy: proof that it means what it says about safety. That credibility is now fueling an $800 billion valuation and partnerships with practically every major tech company. The real test comes when the financial pressure catches up. With tens of billions in investor capital expecting returns, the principled "no" gets harder to say each time, and the next standoff probably won't arrive with such clean optics. But refusing the most powerful institution on earth and watching your revenue explode is, at minimum, a hell of a business strategy.
