OpenAI Pentagon Agreement Triggers QuitGPT Boycott, 1.5 Million Users Exit ChatGPT

OpenAI has reached an agreement with the US Department of Defense to allow its AI models to be used on classified networks, with reports indicating over 1.5 million users have joined the QuitGPT boycott.

The Facts

In late February 2026, OpenAI announced it had signed a contract with the Pentagon to deploy its models on classified networks. The QuitGPT organization claims on its website that as of this week, more than 1.5 million users have canceled their ChatGPT subscriptions, pledged to stop using the service, or shared boycott messages on social media. The development came about a week after Anthropic rejected similar terms, having demanded legal guarantees prohibiting the use of its technology for mass surveillance of US citizens or fully autonomous weapons systems.

Mechanism Breakdown

The agreement permits the government to use OpenAI's technology within legal boundaries, but critics note it could open the door for intelligence agencies to later modify the contract terms. In a post on X, OpenAI's CEO stated that the Department of Defense has demonstrated respect for security and that technical safeguards have been embedded to prevent misuse. Anthropic, for its part, chose to reject the terms and said it would challenge any government action that removes these safeguards through legal channels. The QuitGPT organization has linked the partnership to political donations, pointing out that OpenAI executives contributed $25 million to a Trump-affiliated super PAC.

Industry Impact

In terms of the competitive landscape, Anthropic quickly gained new users after the deal fell through, reaching the top of the Apple App Store charts. OpenAI, meanwhile, faces pressure from user attrition, with its market share already showing a downward trend. Developers need to assess the risks of API dependency — those continuing to use ChatGPT should prepare data export and history deletion procedures, while switching to Anthropic may avoid some ethical controversies but requires confirming service stability. Enterprise users must weigh compliance costs against potential reputational risks when making selection decisions, and government contracts may require additional security reviews.

Comparison and Precedents

This incident stands in direct contrast to Anthropic's rejection of the terms. Anthropic, by holding firm against mass surveillance and autonomous weapons provisions, has been flagged as a supply chain risk, while OpenAI has chosen to fill the void. Historically, cooperation between AI companies and governments has often triggered similar disagreements, but this time the scale of the boycott was rapidly amplified through social media. QuitGPT has collected nearly 20,000 signatures, with a target of 25,600.

Strategic Assessment

The boycott movement could last for several weeks, and OpenAI may seek to narrow the scope of usage by agencies such as the NSA through contract clarifications. The volume of QuitGPT-related discussions on X and changes in Anthropic's download numbers can indicate whether the user migration is stable. Before making actual API calls, developers should test whether the data export functionality is complete to avoid losing historical records. When selecting a provider, enterprises should prioritize comparing the specific safeguard clauses in each company's contract rather than merely looking at market share. OpenAI's current user base is still dominated by younger demographics, and the boycott could further erode this advantage.