QuitGPT protesters rally outside OpenAI’s San Francisco headquarters over Pentagon contract and AI safeguards

Demonstration targets OpenAI’s defense work and broader concerns about surveillance, weapons use, and accountability

Protesters gathered outside OpenAI’s headquarters in San Francisco’s Mission Bay neighborhood this week, rallying under the banner “QuitGPT” in response to the company’s newly announced agreement to provide its artificial intelligence systems for use in classified U.S. Defense Department environments.

Organizers framed the demonstration as opposition to OpenAI’s participation in military and intelligence work, warning that advanced AI models could be used to expand surveillance capabilities and accelerate weapons development. Attendees included a mix of students, educators, technologists, and local political figures. Several participants called for a consumer boycott of ChatGPT and urged employees to oppose the contract internally.

What OpenAI says the agreement allows—and forbids

OpenAI has publicly described the arrangement as a “cloud-only” deployment in classified settings, with the company retaining control over its safety mechanisms and maintaining “in-the-loop” involvement by cleared personnel. OpenAI has also described three red lines it says guide its government work: no mass domestic surveillance, no directing autonomous weapons systems, and no high-stakes automated decisions without human approval.

After the initial announcement drew backlash, OpenAI said it worked with the Defense Department to add language explicitly prohibiting the intentional use of its AI systems for domestic surveillance of U.S. persons, including via commercially acquired personal data. The company also said the agreement does not authorize use by Defense Department intelligence agencies such as the National Security Agency without a separate, new agreement.

• Deployment: cloud-based in classified environments, with OpenAI operating its safety stack
• Stated limits: domestic surveillance prohibition; constraints tied to human control in weapons contexts
• Governance: participation in a proposed working group involving frontier AI labs, cloud providers, and government stakeholders

Why the deal triggered protests now

The demonstration came amid heightened public debate over how rapidly AI tools are being adopted for national security uses and what enforceable boundaries apply when models are deployed inside classified networks. Protesters questioned whether contractual language and technical controls can meaningfully constrain downstream applications, particularly as model capabilities evolve and operational demands change.

Several protesters also raised issues beyond the Pentagon agreement, including concerns about the environmental footprint of large-scale computing, the concentration of power in the AI industry, and the effects of generative AI on creative work and journalism. The presence of these themes reflected a broader coalition dynamic: opposition to military AI use overlapped with wider skepticism about the societal costs of rapid deployment.

OpenAI’s agreement has become a focal point for a larger argument: whether advanced AI systems can be integrated into defense operations without eroding civil liberties or weakening human accountability for the use of force.

What happens next

OpenAI has indicated it will continue refining guardrails for classified deployments while engaging in inter-industry discussions with government partners. Meanwhile, the protest activity suggests continued scrutiny in San Francisco, where past demonstrations have targeted local AI and defense-adjacent firms over militarization, surveillance, and the pace of adoption. How durable the QuitGPT boycott effort becomes may depend on whether the revised contractual terms, oversight mechanisms, and practical implementation satisfy critics who argue that restrictions must be both explicit and enforceable.