OpenAI confirmed an agreement with the U.S. Department of Defense in late February 2026 to deploy advanced AI systems, including ChatGPT technology, on classified military networks. The move immediately sparked widespread controversy, prompting a grassroots #QuitGPT boycott campaign that drove a sharp spike in app uninstalls, a drop in downloads, and a boost for competitors.
#QuitGPT Movement Sweeps the U.S.
According to app analytics data, on Feb. 28, the day of the announcement, ChatGPT uninstall rates in the U.S. surged 295% day over day. Downloads fell 13% the following day and another 5% after that. User sentiment turned sharply negative in app reviews: one-star ratings skyrocketed 775% in a single day and continued climbing, while five-star reviews dropped by roughly half. Competitors capitalized on the moment; Anthropic's Claude app recorded download increases of 37% to 51% during the same period, briefly overtaking ChatGPT in U.S. App Store rankings. Organizers of the boycott claimed millions of actions tied to the campaign, including cancellations and pledges, though exact figures vary depending on the source and participation definition.
OpenAI stressed that the agreement included explicit guardrails: bans on mass domestic surveillance of U.S. persons, autonomous weapons control, and high-stakes automated decision-making systems. It also emphasized technical constraints such as cloud-only deployments, retained control over safety systems, and compliance with U.S. legal frameworks like the Fourth Amendment and Department of Defense rules governing human oversight of lethal force. Despite these safeguards, public concern over AI militarization persisted.
CEO Altman Admits Communication Failures and Revises Terms
Under intensifying public pressure, CEO Sam Altman acknowledged shortcomings in how the deal was communicated, calling the rollout “opportunistic and sloppy.” Within days, the company revised the agreement language: the updated terms explicitly prohibited intentional domestic surveillance using AI systems and added stricter requirements for any intelligence agency involvement, including separate contractual layers. OpenAI also announced plans to coordinate with other AI developers on shared safety frameworks. Altman positioned these changes as a tightening rather than a retreat, which partially calmed the outcry.
Nevertheless, the episode left a mark, highlighting how quickly consumer sentiment can shift when AI crosses into sensitive territory. At the same time, OpenAI was making less visible but strategically significant moves behind the scenes.
Infrastructure Strategy Shift: From Building to Leasing
In early March, OpenAI reorganized its computing and infrastructure operations, splitting responsibilities into three focused groups covering data center design, commercial partnerships, and on-the-ground facility management. The restructuring reflects a broader shift in how OpenAI plans to scale its computing power. Rather than aggressively building and owning massive data centers tied to its ambitious “Stargate” initiative, the company is leaning more heavily on leasing and partnerships with cloud providers. Microsoft’s Azure remains central to that strategy, while OpenAI has also expanded relationships with Oracle and Amazon Web Services as part of multiyear capacity agreements. Earlier plans involving large-scale, jointly owned infrastructure projects have been scaled back or reworked, as the financial and logistical realities of building AI supercomputing capacity at scale become harder to ignore. Instead, OpenAI is focusing on controlling key elements such as custom hardware and chips, while outsourcing the physical infrastructure layer to established hyperscalers.
These two developments — one public and contentious, the other operational and pragmatic — are not directly linked, but together they sketch a company moving quickly on multiple fronts, sometimes faster than its messaging can keep up. For OpenAI, the challenge now is less about whether it can build powerful systems and more about how it manages the consequences of deploying them in places where the stakes are anything but theoretical.