AI already moves fast, but the last few days have been a whirlwind even by its standards. Anthropic refused to give the Pentagon unrestricted access to Claude, after which the Trump administration labeled the company a “supply chain risk” and ordered federal agencies to stop using its technology. Hours later, OpenAI went on to sign the very deal Anthropic had turned down — enabling the Department of War to deploy its AI models across classified military networks.
The funny thing, though, is that Sam Altman, the CEO of OpenAI, had publicly voiced support for Anthropic’s stance just hours before striking its own deal with the Pentagon. Now, OpenAI is in full damage control mode, trying to win back the users it just lost by rewriting the terms of its agreement with the Department of War.
Altman admits the deal was “opportunistic and sloppy”
On Monday, Altman took to X to share that OpenAI has been working with the Department of War to add new language to the agreement and clarify some of the terms. The first point makes it explicitly clear that the AI system “shall not be intentionally used for domestic surveillance of U.S. persons and nationals,” a move meant to reassure the public about civil liberties.
Altman also clarified that the agreement with the Pentagon doesn’t allow OpenAI’s services to automatically extend to intelligence agencies like the NSA. Any use by such agencies would require a separate contract modification, adding another layer of oversight that wasn’t in the original deal.
Within the post, Altman also acknowledged that the company had rushed into the announcement and explained that signing the agreement ended up looking “opportunistic and sloppy,” while their intention was to “de-escalate things and avoid a much worse outcome.” Note that Altman did not address the use of fully autonomous weapons, which happens to be the main sticking point for Anthropic and users alike, alongside mass surveillance.
The damage might already be done
The timing of this amendment says a lot. ChatGPT users were understandably not pleased with the company’s dealings with the Pentagon. A movement called QuitGPT quickly gained traction, with over 1.5 million users pledging to cancel their subscriptions and switch platforms.
Data from Sensor Tower also shows that ChatGPT uninstall rates are up to 295% in the U.S. over the last few days. The same report claims that Claude installs within the U.S. were up 37% on Friday, and then up to 51% last Saturday. Anthropic’s Claude also shot up to the top of Apple’s App Store rankings, replacing ChatGPT at No. 1 for the first time.
It’s worth mentioning here that the QuitGPT movement predates the Pentagon deal. It started in early February 2026 and was sparked by several different reasons. These include OpenAI president Greg Brockman and his wife donating $12.5 million to a pro-Trump supporter (according to public records), ICE using a resume screening tool powered by GPT-4, and a general sense of frustration with GPT-5’s quality.
The Pentagon deal just happened to land at the worst possible time, pouring gasoline on a fire that was already burning.
