After the news of an agreement between OpenAI and the US War Department, uninstallations of the ChatGPT app exploded with a daily increase of 295%, a value that clearly breaks the norm: in the previous weeks, the average removal rate was around 9%. This gap not only signals a decline in popularity of the OpenAI platform, but a conscious stance taken by some users, who have interpreted collaboration with the military apparatus as an outdated ethical line. At the same time, a redistribution of interest towards competing platforms has been observed, in particular towards the ecosystem developed by Anthropicperceived as more prudent in the use of artificial intelligence in sensitive areas.
The CEO of OpenAI, Sam Altman, upon reaching the agreement, wrote in a post on X:
We have reached an agreement with the War Department to deploy our models in their classified network. In all of our interactions, the DoW has demonstrated a deep respect for security and a desire to work together to achieve the best possible outcome.
The mass exodus from ChatGPT towards competing apps demonstrates how little credible the statements made by OpenAI’s number one were judged by users. And, in fact, Altman intervened in a second post, in which he admitted:
We shouldn’t have rushed to publish the news on Friday. The issues are extremely complex and require clear communication.
The agreement between OpenAI and the Pentagon
The reason behind the strong user reaction is therefore attributable to the agreement between OpenAI and the Pentagon. The agreement between the two parties provides for the implementation of OpenAI models in classified military networks. A “classified network” is an IT infrastructure that can only be accessed by authorized parties, because it processes sensitive information for national security. The fear expressed by many users is that systems created for assistance and content generation could be adapted to mass surveillance functions or autonomous weapon systems, i.e. capable of selecting and hitting targets without direct human control. In response to these concerns, OpenAI CEO Sam Altman said he would make some changes to the original agreement:
We have worked with the Department of Defense to make some additions to our agreement to clarify our principles. (…) There are many things that the technology is not yet ready for and many areas where we do not yet understand the trade-offs necessary for security. We will work on these aspects, slowly, with the DoW, with technical guarantees and other methods.
Privacy concerns
A central component of the negative reaction recorded towards ChatGPT concerns privacy and the fear that artificial intelligence could be used for mass surveillance or in military contexts without adequate ethical guarantees. The “QuitGPT” campaign, for example, claims that OpenAI would have agreed to allow the Pentagon to use the technology to «any lawful purpose» included «mass surveillance of Americans» and the «producing AI-based weapons that kill without human supervision». These are very strong statements, which evidently contributed to ChatGPT’s decline in popularity. Many users took action not only by deleting the app, but also by sharing negative reviews: according to the Sensor Tower platform, one-star ratings for ChatGPT grew by 775% on Saturday and 100% on Sunday, while five-star ratings decreased by 50% in the same period.
Downloads growing for Claude by Anthropic
In parallel with the decrease in ChatGPT downloads, there has been an increase in interest in Claude, the chatbot developed by Anthropic. This company has publicly stated that it rejected a request from the War Department for unrestricted access to its systems, citing ethical concerns about possible military or surveillance uses that its executives say current AI cannot safely handle. Dario Amodei, CEO of Anthropic, explained:
In a limited number of cases, we believe AI can undermine, rather than defend, democratic values. Some uses are simply outside of what today’s technology can do safely and reliably.
Anthropic’s stance had concrete consequences: according to Appfigures, a well-known service that tracks statistics on software applications, for the first time Claude’s daily downloads in the United States exceeded those of ChatGPT, with an estimated jump of up to 88%. The app topped the US App Store chart and also reached first place in several European countries.
