OpenAI just confirmed a deal with the Pentagon to deploy its AI models in classified settings, following failed talks between the military and Anthropic. The agreement was, in Sam Altman’s words, “definitely rushed” and has already triggered backlash across tech circles.
This move matters for anyone watching the intersection of AI and government power. Anthropic refused to cross certain lines-like enabling autonomous weapons or mass domestic surveillance. OpenAI claims to share those red lines, but critics aren’t convinced the contract language is airtight.
What OpenAI Says-And What Critics See
In a blog post, OpenAI outlined three hard stops for its tech: no mass domestic surveillance, no autonomous weapons, and no high-stakes automated decisions (think social credit systems). The company insists its “multi-layered approach”-cloud-only deployment, cleared personnel, and contractual protections-keeps things safe. “We retain full discretion over our safety stack,” the blog claims.
But Mike Masnick from Techdirt called out a loophole: the deal allows data collection under Executive Order 12333. That order, he says, lets the NSA scoop up communications outside the US-even if Americans are involved. Masnick argues this is just a legal backdoor for domestic surveillance, no matter what the contract says.
Katrina Mulligan, OpenAI’s head of national security partnerships, pushed back on LinkedIn. She argued that deployment architecture matters more than contract language. By keeping everything on the cloud API, she claims, OpenAI’s models can’t be directly integrated into weapons or operational hardware.
Why the Rush?
Sam Altman admitted on X that the deal was rushed and the optics are ugly. He said the goal was to “de-escalate things” between the Department of Defense and the AI industry. But the backlash was swift-so much so that Anthropic’s Claude overtook OpenAI’s ChatGPT in Apple’s App Store right after the news broke.
Meanwhile, President Donald Trump ordered federal agencies to stop using Anthropic’s tech within six months, and Secretary of Defense Pete Hegseth labeled the company a supply-chain risk. That left OpenAI as the Pentagon’s AI partner, at least for now.
The bottom line
- OpenAI’s Pentagon deal is live, but critics say it could enable domestic surveillance through legal loopholes.
- Gamers and tech users should watch for real-world impacts-especially if AI policy shifts or government use expands.
Speculation: If OpenAI’s safeguards hold, it could set a new standard for AI in national security. If not, expect more scrutiny-and maybe a player exodus to rivals like Anthropic.