OpenAI’s decision to enter into an agreement with the United States Department of Defense had an immediate effect on user behavior. Data from market intelligence companies shows that shortly after the partnership was announced, ChatGPT’s mobile app saw a sharp drop in uninstalls, while competitors like Anthropic’s Claude gained traction.
According to Sensor Tower, ChatGPT uninstalls in the United States grew 295% in a single day, on Saturday, February 28th. For comparison purposes, the average daily app removal rate had been around 9% in the previous 30 days. The movement occurred after the agreement between OpenAI and the Department of Defense, renamed by the Trump administration as Department of War, became public.
In addition to the increase in uninstallations, ChatGPT downloads also felt the impact. On the same Saturday, app installations in the US fell 13% compared to the previous day, a decline that deepened on Sunday, with an additional drop of 5%. The contrast draws attention because, on the Friday before the agreement was announced, the application had recorded 14% growth in downloads.
User dissatisfaction also appeared in ChatGPT reviews. Reviews with just one star grew 775% on Saturday and increased by another 100% on Sunday. In the same period, five-star reviews fell by 50%, according to Sensor Tower.
Competitor takes the lead
In the opposite direction, Claude showed strong acceleration. Also according to Sensor Tower, downloads of the Anthropic app grew 37% on Friday, February 27, and 51% on Saturday. The increase came after the company made public that it had refused a partnership with the Department of Defense because it did not agree to the terms, citing concerns about the use of AI for citizen surveillance and fully autonomous weapons.
The data suggests that some consumers reacted positively to Anthropic’s stance. The movement was enough to take Claude to the top of the App Store in the United States: on Saturday, the application reached first position in the ranking and remained there until at least the following Monday, March 2nd. About a week earlier, on February 22, the app occupied a position more than 20 places lower.
Continues after advertising
Data company Appfigures pointed out that, for the first time, daily downloads of Claude in the US surpassed those of ChatGPT. According to the company’s estimates, installations of the Anthropic app grew 88% in a single day. Claude now leads the ranking of free iPhone apps in six countries outside the USA, including Belgium, Canada, Germany, Luxembourg, Norway and Switzerland.
Similarweb noted that Claude’s downloads in the United States last week were around 20 times higher than in January. The company, however, considers that the jump may be linked to factors other than political issues surrounding the military agreement.
Ahurried chord
The negative reaction occurs in a delicate political context. After negotiations between Anthropic and the Pentagon broke down in late February, President Donald Trump ordered federal agencies to stop using the startup’s technology after a six-month transition period. Defense Secretary Pete Hegseth even classified the company as a risk to the supply chain.
Continues after advertising
Shortly afterwards, OpenAI announced that it had reached its own agreement, providing for the use of the company’s models in classified environments. The CEO, Sam Altman, publicly acknowledged that the process was “definitely rushed” and that “the perception was not good”, admitting the reputational damage.
To respond to the criticism, OpenAI published a post detailing its safeguards. The company says its models cannot be used for mass domestic surveillance, autonomous weapons systems, or high-impact automated decisions such as “social credit” systems. According to the company, the agreement protects these restrictions through a “broader and multi-layered” approach, including full control over its security infrastructure, use via the cloud, participation of OpenAI’s own authorized personnel and contractual protections.
Open debate
The explanations, however, did not end the debate. After the publication of the OpenAI post, Mike Masnick, from the blog Techdirt, stated that the contract “absolutely allows domestic surveillance”, citing the reference to Executive Order 12333, which regulates the collection of intelligence data outside the United States, even when it involves information from American citizens.
Continues after advertising
In defense of the startup, Katrina Mulligan, responsible for national security partnerships at OpenAI, argued in a LinkedIn post that the debate has overestimated the weight of isolated contractual clauses. According to her, the deployment architecture is more relevant than the contract text, and the exclusive use of cloud APIs would prevent direct integration of the models with weapons, sensors or other operating systems.
Sam Altman, in turn, stated that OpenAI decided to proceed with the agreement in an attempt to “de-escalate” tensions between the government and the AI industry. For him, the reputational risk could be worth it if the move helps establish a new balance. Otherwise, he admitted, the company will continue to be seen as hasty and not very careful – a perception that, at least for now, is already reflected in the app store numbers.
Content produced by .
