OpenAI’s Sam Altman announces Pentagon deal with ‘technical safeguards’
The agreement includes technical guardrails against mass surveillance and autonomous weapons, a key point of contention.
OpenAI has finalized a landmark agreement with the U.S. Department of Defense, granting the Pentagon access to its AI models for use on classified networks. The deal, announced by CEO Sam Altman, comes after a high-profile breakdown in negotiations between the DoD and AI rival Anthropic, which had refused to allow its models to be used for "all lawful purposes" over ethical concerns. The contentious standoff saw over 60 OpenAI and 300 Google employees sign an open letter supporting Anthropic's position, while former President Trump criticized the company and the Pentagon designated it a supply-chain risk.
A critical component of OpenAI's deal is the inclusion of specific "technical safeguards" aligned with the company's safety principles. Altman stated the contract explicitly prohibits uses for domestic mass surveillance and ensures human responsibility for the use of force, addressing the core ethical flashpoints that derailed the Anthropic talks. OpenAI will deploy engineers to work with the Pentagon and build a proprietary "safety stack," with the government agreeing not to force the company to override a model's refusal to perform a task. This agreement sets a new precedent for military-AI partnerships, balancing operational access with developer-enforced ethical boundaries.
- Deal includes explicit bans on use for domestic mass surveillance and fully autonomous weapon systems
- Follows a failed DoD negotiation with Anthropic that led to a federal supply-chain risk designation
- OpenAI will build a proprietary "safety stack" and deploy engineers to ensure model compliance
Why It Matters
Sets a major precedent for how AI companies can engage with military clients while enforcing their own ethical guardrails.