Model Choice with a Stance:
Why Claude Became the #1 App in the US App Store Overnight

In February 2026, Claude became the most-downloaded AI app in the US overnight — not because of marketing, but because of a stance. When the Pentagon demanded softer usage terms, Anthropic refused. The result: a court case, a user rebellion against OpenAI — and a fundamental question for every company that uses AI.

What happened? The Pentagon deal and the backlash

On February 28, 2026, OpenAI announced an agreement with the US Department of Defense: AI models on classified Pentagon networks, delivered via a cloud-based API. Purpose: intelligence analysis, decision support, cybersecurity. What was meant as a strategic partnership triggered a backlash within hours.

A few days later, Sam Altman admitted the deal had been "opportunistic and sloppy" (CNBC, 03/03/2026). OpenAI subsequently tightened its red lines: no mass domestic surveillance, no control of autonomous weapons systems, and the NSA explicitly excluded.

The user reaction was measurable. According to Sensor Tower, daily ChatGPT uninstalls rose 295% above average on February 28. One-star App Store ratings jumped 775%. Forbes reported that around 1.5 million users left ChatGPT in the immediate aftermath. The #QuitGPT campaign, which had already existed since February 10 (MIT Technology Review), became a cross-platform phenomenon because of the Pentagon deal: X/Twitter (trending globally for three days), TikTok (cancellation videos), LinkedIn (B2B discussion).

How did Anthropic respond? Refusing to soften its AUP

Anthropic had already been running Claude on classified Pentagon networks since July 2025 — the first frontier model ever to do so. The contract contained a clear commitment from the Pentagon: use in accordance with Anthropic's Acceptable Use Policy (AUP), which rules out mass surveillance and fully autonomous weapons systems.

In the first quarter of 2026, the Pentagon demanded these clauses be struck — Claude was to become usable "for all lawful purposes." DoD Under Secretary Emil Michael called on Anthropic to "cross the Rubicon." The deadline expired on February 27, 2026, at 5:01 p.m. Anthropic did not agree.

The consequences followed quickly: an executive order for a phase-out with a 180-day deadline, classification as a "supply chain risk," and the termination of a contract worth around $200 million. Anthropic sued on March 9, 2026 — and won a preliminary injunction on March 26 over "First Amendment retaliation." The appeal is ongoing: the appellate hearing took place in May 2026, and according to Federal News Network the judges were divided. The ruling is still pending.

Why did Claude become the #1 app in the App Store?

On the evening of February 28, 2026, Claude reached #1 among the top free apps in the US App Store — the first time since ChatGPT's launch in November 2022 that another AI assistant took that position. Sensor Tower documents the trajectory: outside the top 100 at the end of January, top 20 in February, then #1 within four weeks.

Claude downloads rose 37% on February 28 and 51% on March 1. Anthropic reported: daily sign-ups at an all-time high, free users up 60% since January, paying subscribers more than doubled year-to-date.

What does model choice mean for mid-sized businesses?

For companies that deploy AI agents in customer-facing roles — whether a phone agent, a recruiting assistant, or a dunning bot — their model provider's Acceptable Use Policy isn't fine print. It's the chain of accountability.

If you operate an AI agent that processes GDPR-relevant data, you concretely benefit when your model provider enforces its usage terms instead of softening them under pressure. The EU AI Act classifies systems by risk — and a provider's stance is a factor that regulators increasingly weigh.

At Amazing Agents, we build on Claude — not because Anthropic is perfect, but because the chain of accountability stays traceable. We choose models based on accountability, not price.

While you're reading this, a ruling is coming down

The Anthropic vs. Pentagon appeal isn't over. What began as a #QuitGPT wave in the spring is now being decided in the federal court of appeals. Regardless of the outcome, the case shows: model choice is no longer a purely technical decision. It's a statement about what values your company carries through its tools.

Want to understand what agentic AI means for your business? Our Agentic AI Guide explains the fundamentals. And if you want to hear what an Anthropic-powered AI agent sounds like in customer contact: meet Ben.