How Anthropic’s Pentagon fight became an unexpected growth engine

A version of this article was originally published in Quartz’s AI & Tech newsletter. Sign up here to receive the latest news, analysis and insights on AI and technology straight to your inbox.

When the Pentagon declared Anthropic as a supply chain risk Earlier this year, this designation was supposed to hurt. That label, usually reserved for antagonistic foreign companies, was a pointed message from Defense Secretary Pete Hegseth after Anthropic refused to let the military use its AI models for autonomous weapons or domestic surveillance. Anthropic called the move “legally unsound” and filed a lawsuit.

What followed was less of a corporate crisis than a coming out party.

In the months since the standoff began, Anthropic says its annualized revenue has fallen from about $9 billion at the end of 2025 to more than 30 billion dollars todayand consumers’ paid subscriptions more than doubled. The Claude application briefly in the lead Apple $AAPL download charts.

The legal battle has temporarily produced a split decision — a court blocked the government from imposing a ban on Claude, while a federal appeals court allowed the Pentagon’s blacklist to remain while the litigation plays out.

And last week, Anthropic unveiled Project Glasswing, a broad cybersecurity initiative that brought together partners including AWS, Apple, Microsoft $MSFT, Google $GOOGL, and Cisco $CSCO to test a new, unique model called Claude Mythos – described in leaked internal documents as “by far the most powerful AI model we have ever developed.”

For a company that spent years in OpenAI’s shadow, it’s been a remarkable few months.

The feud turned into a marketing campaign

The conflict at the Pentagon gave Anthropic something that money rarely buys for an AI company: a clear moral identity. As OpenAI partnered with the Department of Defense, Anthropic sued it. The contrast was stark and consumers noticed it.

When OpenAI announced its own deal with the Pentagon, ChatGPT uninstalls jumped 295% day over dayaccording to market information company Sensor Tower. Claude’s downloads increased by 51% over the same weekend.

An analysis of credit card data from approximately 28 million U.S. consumers found that new paying subscribers have increased significantly in the weeks between the first reports of the impasse and CEO Dario Amodei’s public statement on the matter in late February. Previous users returned to the platform in record numbers in the same month.

However, revenue growth is first and foremost a business story. The number of customers spending at least $1 million per year has more than doubled since February, cross 1,000 customersaccording to Anthropic.

Claude Code, the company development tool released in Januaryhas been a major driver, rapidly adding subscribers and helping to push run rate past milestones that took other software companies decades to reach.

A model too powerful to go out

The Glasswing project arrived as a sort of cornerstone of the sequence. The initiative is structured around Claude Mythe Overviewa model that Anthropic claims can find and exploit hidden software vulnerabilities at a level that surpasses almost any human security expert.

The company has committed $100 million to allow its partners to use it for defensive security purposes and says it has no plans to make it public. This is a remarkable position for a company whose business depends on the people using its models. The implication is that Mythos is simply too capable to freely hand over the baton.

The announcement had a dual function. It demonstrated that Anthropic’s security-focused positioning is more than rhetoric, and presented a model that the company has explicitly positioned as too dangerous for large-scale deployment.

This formulation aroused some skepticism. Talking about the risks of a model is a well-known approach in the AI ​​industry, and the timing, coinciding with IPO discussion reports for later this year, has not escaped observers.

However, the partners involved are not small names. Microsoft, Google, Crowd strike $CRWD and JPMorganChase have all reported using Mythos Preview in their own security operations. The credibility that comes with these endorsements is harder to manufacture than a press release.

The Pentagon, for its part, is not giving up. The appeals court’s ruling means Anthropic remains barred from Defense Department contracts for now, although it can continue to work with other government agencies. Emil Michael, Undersecretary of Defense continued to publicly attack Amodeiand the legal battle remains unresolved. But Anthropic continues to grow.

Meanwhile, OpenAI is taking some hits. A recent New York profile CEO Sam Altman drew on accounts from former colleagues describing him as slippery at best. The piece came as OpenAI also reportedly faced internal friction with its CFO ahead of its own planned IPO. And in March, The Wall Street Journal reported that top executives were finalizing plans to move away from “side quests,” including its video tool Sora, to refocus on coding and enterprise customers, the exact ground Anthropic has spent the last year consolidating.

None of this is Anthropic’s doing. But it’s hard not to notice the contrast with Amodei’s public positioning.