Some refusals come at a steep price; others yield hefty rewards. Anthropic has just experienced the latter—on the battlefield of popularity.
By walking away from a contract with the U.S. Department of Defense—one that OpenAI, its rival, accepted in its wake—the scale-up led by Dario Amodei vaulted to the top of the American App Store, surpassing ChatGPT for the first time in its history.
Everything began in early 2026, when the Pentagon sought to deploy AI models across its classified networks. Anthropic stayed in the race until the moment it realized the deal would require lifting some of its “red lines.” Not a chance for Dario Amodei and his team to permit domestic surveillance of American citizens or to endorse the development of lethal autonomous weapons. Anthropic slammed the door. OpenAI walked in.
A Botched Negotiation, According to Sam Altman
OpenAI’s decision triggers a chain reaction that few had anticipated. In San Francisco as well as in London, protesters pitched tents outside OpenAI’s offices, center stage for Sam Altman’s company.
But the most unexpected blow came from within: nearly 500 employees from OpenAI and Google co-sign a public letter titled “We Will Not Be Divided,” expressing their support for Anthropic’s positions. Faced with the storm, OpenAI’s CEO is compelled to make a rare public self-critique, admitting that the negotiation had been “botched,” and promising urgent amendments to the contract.
These amendments, hammered out under pressure, now include explicit prohibitions: AI systems “must not be used intentionally for domestic surveillance” of American citizens.
Intelligence agencies such as the NSA are also barred from access without prior contract changes, a guarantee that the Pentagon itself publicly confirmed. Too late, in the eyes of many observers.
OpenAI vs Anthropic: a Clash of Principles
The growth has been spectacular. Claude has become the most downloaded free AI app on the American App Store, pushing ChatGPT to second place and placing Gemini further behind. On Android, the app climbs to seventh overall. Daily signup records are being set one after another, and paid subscribers have doubled in the span of a year.
The movement goes beyond the circle of discerning users. The QuitGPT group is actively urging a boycott of ChatGPT on social networks, encouraging people to uninstall the app. During the latest Super Bowl, Anthropic aired a 60-second commercial that drew intense commentary, poking fun at the growing integration of sponsored content within ChatGPT.
OpenAI Under Pressure
At OpenAI, renegotiation is now conducted under constraint. The amendments announced by Sam Altman read less like a principled stand and more like crisis management. Still, the underlying question will recur: how far can civil AI giants commit to military applications without losing the public’s trust?
The controversy also sparks a broader debate about regulating military AIs. Fears of lethal autonomous weapons and of “social credit” systems inspired by the Chinese model fuel an international movement that goes beyond a mere squabble between two AI heavyweights. Europe, already engaged in drafting its AI Act, is watching carefully.
For Anthropic, the challenge is now to convert this momentum into a durable advantage. The scale-up remains unprofitable, in a race to build out infrastructure that requires billions of dollars in investment. But it has demonstrated one thing that few of its rivals would have dared to bet on: in the attention economy of 2026, adhering to principles can be profitable.
