The dust has barely settled on OpenAI’s hastily arranged handshake with the Pentagon, but the shockwaves are still radiating out from San Francisco to Washington. What began as a routine government contract swap has exploded into a full-blown existential crisis for the ChatGPT maker, marked by a user revolt, high-profile resignations, and a rare public admission of error from its normally unflappable CEO, Sam Altman.

Just over a week ago, the Department of Defense (DoD) found itself in need of a new artificial intelligence partner. Its previous contractor, Anthropic the maker of the Claude AI model and a company built on a foundation of rigorous safety protocols, was shown the door. The reason? According to Defense Secretary Pete Hegseth, Anthropic’s insistence on contractual “red lines” against using its technology for mass domestic surveillance or in fully autonomous weapons systems was a non-starter. Hegseth labeled the company “left-wing nut jobs” and ordered an immediate divorce.

Into that vacuum stepped OpenAI. Within hours of Anthropic being blacklisted on February 27, the company had a new deal in place to supply its AI for classified military operations. It was a swift, decisive power play. But as Altman himself would later admit, it looked exactly like that: a company swooping in to capitalize on a rival’s misfortune without thinking through the consequences. “We shouldn’t have rushed to get this out on Friday,” he wrote in a memo to employees. “The issues are super complex, and demand clear communication. We were genuinely trying to de-escalate things, but I think it just looked opportunistic and sloppy”.

That moment of candor, however, came too late to stop a cascade of backlash that is now threatening to define the company’s legacy.

A QuitGPT Movement and the Flight to Claude

The immediate response from the user base was visceral. Overnight, OpenAI went from being the plucky creator of the world’s most accessible chatbot to being branded a “war machine” by its own customers. On X (formerly Twitter) and Reddit, users launched what analysts are calling a “QuitGPT” campaign, encouraging mass cancellations of ChatGPT subscriptions.

The numbers tell a stark story of consumer rage. Data from Sensor Tower showed that uninstalls of the ChatGPT mobile app surged by a staggering 295% on the Saturday following the announcement, a massive spike compared to the usual daily average of 9%. Reports from Chinese media outlet Sina suggested that over 700,000 users may have pulled the plug on their subscriptions in a matter of days.

In a delicious irony for Anthropic, the very company ousted by the Pentagon for being too ethical, the backlash has become a business boon. As users fled OpenAI, they flocked to Claude. The Anthropic app rocketed to the top of Apple’s App Store charts, leapfrogging its rival ChatGPT for the first time in recent memory. For many, downloading Claude was not just about finding a new AI tool; it was a political statement—a vote against the militarization of the technology they use every day.

Silicon Valley’s Revolt: “Principle, Not People”

While the user revolt was damaging, the bleeding soon turned internal. The deal triggered a crisis of conscience among the very people building the technology.

The first major crack appeared when nearly 900 employees from both OpenAI and Google signed an open letter, pleading with their leadership to stand firm against the DoD’s demands. They warned that the government was trying to “divide each company with fear that the other will give in,” and they explicitly refused to let their work be used for “domestic mass surveillance and autonomously killing people without human oversight”.

But the rhetoric became reality on March 6, when Caitlin Kalinowski, OpenAI’s head of robotics, walked out the door. In a resignation post on X that has since gone viral, Kalinowski made it clear she wasn’t leaving due to personality conflicts or a better offer. She left on principle. “This wasn’t an easy call,” she wrote. “AI has an important role in national security. But surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got”.

Kalinowski, a veteran who previously built hardware at Meta and Apple, did not mince words about the company’s governance. She stated that the deal was “rushed without the guardrails defined,” adding, “It’s a governance concern first and foremost”. Her departure serves as a powerful symbol that, for some insiders, the “sloppy” execution wasn’t just a PR problem, but a moral breach.

 Altman’s Awkward Walk Back

Facing a PR wildfire and a staff exodus, Sam Altman was forced into damage control mode. On Monday, he announced that the company would be amending the deal, which had been signed just days prior .

The revisions were an attempt to patch the holes Anthropic had identified weeks earlier. The amended contract now explicitly prohibits the DoD from using OpenAI’s technology for “domestic surveillance of U.S. persons and nationals,” citing compliance with the Fourth Amendment. Altman further clarified that the agreement would not automatically extend to intelligence agencies like the National Security Agency (NSA) without a “follow-on modification” to the contract.

Yet, for many critics, these amendments feel like a case of too little, too late. As MIT Technology Review pointed out in a scathing analysis, OpenAI’s “compromise” relies on the assumption that the government will simply follow the law. But as the Edward Snowden revelations proved, the government’s interpretation of what is “lawful” can sometimes diverge wildly from the public’s expectation of privacy.

The company’s defense is that it can embed its “red lines” into the model’s behavior, refusing to answer queries related to surveillance. However, experts remain skeptical. “With Anthropic out of the Pentagon, the most safety-conscious actor is now out of the room,” Professor Mariarosaria Taddeo of Oxford University told the BBC. “That is a real problem”.

As the U.S. government pushes for new guidelines that would force AI companies to allow “all lawful uses” of their models, OpenAI finds itself stuck between a rock and a hard place. By trying to please Washington, it has alienated its users and its talent. And by trying to please the public with hastily added “guardrails,” it risks looking like a company that doesn’t know what it stands for. The fallout is far from over.