OpenAI renegotiates “opportunistic, sloppy” Pentagon deal, sparking ChatGPT boycott debate

OpenAI’s revised Pentagon contract now bars any intentional domestic surveillance use of its models, a clause Altman said was added after “rushed” negotiations, according to Fortune. The amendment cites the Fourth Amendment, the National Security Act of 1947 and the Foreign Intelligence Surveillance Act of 1978, and explicitly prohibits the Defense Intelligence Agency, NSA, NGA and other intelligence components from accessing the service without a separate modification (Fortune).

The change comes as a “QuitGPT” boycott gains traction. Euronews reports more than 1.5 million users have either canceled subscriptions, posted boycott messages or signed up at quitgpt.org since the deal’s disclosure. The campaign frames the Pentagon contract as a breach of democratic norms, echoing Anthropic CEO Dario Amodei’s refusal to grant unrestricted military access, which he called “incompatible with democratic values” (Euronews).

Anthropic’s stance has pressured the DoD to tighten its own terms. The Atlantic notes Anthropic demanded guarantees against domestic surveillance before signing its $200 million contract, a demand now mirrored in OpenAI’s renegotiated language (The Atlantic). Katrina Mulligan, OpenAI’s head of national security partnerships and former Pentagon, NSC and DOJ official, confirmed that any future use of OpenAI tools by the listed intelligence agencies will require a distinct contract amendment (Fortune).

Industry observers see the renegotiation as a litmus test for AI firms’ ability to set red lines with the military. The Verge has highlighted similar disputes, reporting that Anthropic is in “heated negotiations” after refusing to loosen guardrails that would enable lethal autonomous weapons or mass data collection (The Verge). OpenAI’s move may signal a broader shift, but the boycott’s momentum suggests public scrutiny will remain intense.

The revised agreement also closes a legal gray area around commercially sourced data. Fortune says the new terms forbid the use of cell‑phone location records, fitness‑app data and other purchased datasets for domestic surveillance, aligning OpenAI’s policy with existing privacy statutes. If the Pentagon proceeds under the amended contract, it will have to seek separate approvals for any intelligence‑community use, a procedural hurdle that could slow deployment of AI‑driven analytics in classified networks.

The backlash underscores a growing tension between national‑security imperatives and civil‑rights concerns. As the DoD pushes for AI integration, firms like OpenAI and Anthropic are being forced to choose between lucrative contracts and the ethical constraints demanded by their user bases. The outcome of these negotiations will likely shape the future of AI in defense and set precedents for how private tech companies negotiate with government agencies.