QuitGPT: How a Pentagon deal became OpenAI’s biggest PR crisis

On a Tuesday evening earlier this month, a 26-year-old Oakland tech worker stood outside OpenAI's headquarters in San Francisco wearing a cardboard robot mask fitted with red LED lights.

He had built it the previous night using a box, black duct tape, and $12 worth of blinking lights. "I bet a lot more people are gonna pay attention to this than OpenAI's next million-dollar ad," Business Insider quoted him as saying.

He was not wrong, at least not entirely. The protest he joined drew many people. Chalk slogans covered the pavement. Signs read "OpenAI, there is blood on your hands." 

It was modest in scale but striking in mood. And behind it sat a movement that, in a matter of days, pulled in more than a million people worldwide.

A deal that broke the dam

The memory of the Cambridge Analytica scandal still haunts American people. 

In 2018, it was revealed that the consulting firm harvested data from millions of Facebook users to help the Trump campaign target voters with remarkable psychological precision. Facebook was the era's defining tech platform, used daily by billions around the world. Today, OpenAI sits in a comparable position.

This time, the trigger was the Pentagon contract.

On 27 February, Anthropic, the company behind the Claude chatbot, refused to grant the US Department of War unrestricted access to its artificial intelligence systems.

The sticking points were two: mass domestic surveillance and fully autonomous weapons. Anthropic drew a line. The Trump administration responded by ordering federal agencies to stop using Claude and, according to Anthropic, threatened to label the company a "supply-chain risk to national security" — a designation normally reserved for Chinese firms such as Huawei.

Within hours, OpenAI's chief executive Sam Altman signed a deal to take Anthropic's place. The fallout starts right there. Calls to cancel ChatGPT subscriptions spread across social media. 'QuitGPT' campaign snowballed to gain dizzying attention.

Claude became number one in the US App Store. It surpassed ChatGPT in daily American downloads for the first time, according to analytics firm Appfigures.

Altman later admitted on X that the rollout had been "opportunistic and sloppy" and said the company was revising its deal with the Pentagon to include explicit protections. The Pentagon did not respond to Business Insider's questions about the amended agreement.

Why a boycott works in this case

Dutch historian and author Rutger Bregman, who has offered an analysis of what QuitGPT might actually become, wrote in The Guardian, "OpenAI is on track to lose $14 billion this year." He observed that the tech giant's market share had "plummeted from 69% to 45% in a single year".

In that vulnerability, Bregman sees an opening.

He draws a deliberate parallel to the 1955 Montgomery bus boycott. Black residents of Alabama, he argued, did not attempt to dismantle segregation all at once. They picked one target and held firm for 381 days until the bus company broke. 

"OpenAI is our bus company," Bregman wrote. What makes QuitGPT potentially effective, in his view, is narrowness and ease, the same thing that made earlier boycotts work. 

The 1977 Nestlé boycott succeeded because there were other tins of formula on the shelf. The 2023 Bud Light boycott succeeded because the market had other options.

Cancelling a ChatGPT subscription, Bregman argues, takes 10 seconds and leaves the user with several credible alternatives. 

"The great boycotts of history did not succeed because millions of people suddenly became heroic activists," he wrote. "They succeeded because buying a different brand of coffee was something anyone could do on a Tuesday afternoon."

More than one grievance

At the protest outside OpenAI headquarters, Business Insiders' Katherine Li found a crowd with multiple and sometimes competing grievances.

Climate activist Perrin Milliken was there to oppose data centres and their water consumption. "AI is taking water from communities, polluting communities, and increasing their electricity bills," she said. "They are not even paying for it. We are."

Protest attendee Sarah Gao turned her attention to wealth inequality and Altman's San Francisco mansion. Some attendees had rejected artificial intelligence entirely on principle. One woman told Li that she could "see exactly where it heads" from the moment it appeared in visual media. "It destroys journalism, it destroys art," she said.

The tech worker in the robot mask occupied a different position from all of them. He uses artificial intelligence every day. He is not against the technology, rather he is against the unethical uses. 

"What I do not want is for the technologies that my friends and I build to be used to undermine the freedom we value," he said.

That tension, between the utility of the tool and the conduct of the company that builds it, is what makes QuitGPT something other than a simple protest. And we are yet to see whether the protests and cancelled subscriptions can add up to a political earthquake.

Nonetheless, the cracks in ChatGPT's dominance are already visible.