ChatGPT’s Pentagon deal sparks user exodus — and a boom for Claude 🧠

Cliquez ici pour lire en français

Since early March, downloads of the AI chatbot have dropped 13% — while user reviews have surged by 775%.

From buzz to backlash 😊 ☹️

ChatGPT has a trust problem. With over 400 million users worldwide, OpenAI’s flagship product has long benefited from near-universal adoption — but most of those users have little control over what happens to their data. That tension came to a head on February 28, when OpenAI’s partnership with the U.S. Department of Defense was made public.

Under normal circumstances, the app’s uninstall rate hovers around 9%. In the three days following the announcement, that figure spiked by 295%. The backlash was all the more striking given that the deal had initially sparked a 14% uptick in downloads — enthusiasm that quickly curdled into concern. Users’ main worries center on data privacy and what level of confidentiality, if any, would be maintained under the agreement.

Sam Altman on the defensive 😨

OpenAI CEO Sam Altman was forced to address the fallout head-on. Speaking to CNBC, he acknowledged that the deal had been struck in an « opportunistic and sloppy » manner. The company has since pledged to revise the original contract terms, particularly around mass surveillance capabilities and the use of autonomous weapons systems.

Meanwhile, OpenAI employees didn’t stay silent either. A group of staff members sent an open letter condemning the agreement, citing serious concerns over data privacy and warning that the technology could be weaponized in « autonomous, unsupervised » systems — a particularly charged issue given the number of active conflicts around the world.

Claude steps into the spotlight 🛡️

The controversy has been a windfall for the competition. Faced with growing unease over ChatGPT, many users pivoted to Claude, the rival AI developed by Anthropic. Within just 24 hours of the news breaking, Claude’s downloads jumped 37%, then climbed another 51% — briefly overtaking ChatGPT on the App Store charts, according to Appfigures.

Sensor Tower data backs this up: Claude has now pulled ahead of ChatGPT in six countries — Canada, Belgium, Switzerland, Luxembourg, Norway, and Germany.

Much of Claude’s appeal in this moment comes down to Anthropic’s firm public stance. The company has made clear it will not hand over bulk user data or allow its AI to be used for military purposes. As CEO Dario Amodei put it: « We cannot in good conscience comply with their request. (…) In some cases, we believe AI can undermine, rather than uphold, democratic values. »

The episode is a sharp reminder that data privacy and user trust aren’t just compliance checkboxes — they’re the foundation on which the AI industry’s credibility rests. And right now, that foundation is being tested.

So, how much do you actually trust AI? Drop your thoughts in the comments.