OpenAI, Pentagon revise AI deal to restrict intelligence agency use

OpenAI said on March 2 that it is working with the US Department of Defense to amend its agreement to clarify how the company’s services may be used, CEO Sam Altman wrote on X.

Altman said one amendment stated the Pentagon affirmed OpenAI services would not be used by the Defense Department’s intelligence agencies such as the NSA without a follow-on contract modification.

The disclosure came after OpenAI last week announced a deal to deploy its technology on the Defense Department’s classified network.

Altman’s post did not specify the scope or timing of the contractual changes, Reuters said.

Reuters’ report appeared to contain a typo referring to a “Department of War”; Altman’s post referenced the Department of Defense (DoD).

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

OpenAI’s contract amendment follows public and internal backlash

The original Defense Department deal landed hours after the Trump administration blacklisted Anthropic, after it refused Pentagon terms that allowed “all lawful purposes” and asked for clear bans on mass domestic surveillance and fully autonomous weapons 1.
OpenAI first agreed to the “all lawful purposes” wording, which critics said could allow surveillance because some laws have been read that way before 2.
The deal set off protests outside OpenAI’s offices, user campaigns urging a switch to rival tools, plus open letters signed by hundreds of employees at OpenAI and Google 3.
CEO Sam Altman later said the rollout “looked opportunistic and sloppy,” adding that the company rushed the announcement while trying to ease government pressure on the AI industry 4.

The Pentagon’s AI push is opening a new rift for tech giants

By labeling a hesitant firm like Anthropic a “supply-chain risk” while signing with OpenAI, the government gives AI labs a strong reason to relax ethical limits 5.
The supply-chain risk tag raises stakes for tech vendors, since Defense Department suppliers may have to certify they do not use Anthropic’s models 1.
The dispute has energized tech workers, with hundreds backing Anthropic in letters that link military deals to recruiting and retention 6.
The moment pressures the AI field to decide whether voluntary ethics can hold under government demands, or whether legal permission will set the boundary instead of what labs consider safe 7.