It feels like just yesterday that OpenAI was the company vowing to keep its powerful AI out of the military's hands. Now, in a pivot that gave a lot of us in the tech world whiplash, they've struck a deal to let the Pentagon use their technology in classified settings.
It’s a huge move, and frankly, it raises a ton of questions. Sam Altman insists his company’s AI won’t be used to build autonomous weapons, but the agreement basically just says the military has to follow its own rules—which are pretty flexible, to say the least.
So, why the sudden change of heart? Maybe it’s just about the money. Training these massive AI models costs an absolute fortune, and OpenAI is burning through cash. Or maybe, as Altman often suggests, he genuinely believes that for the U.S. to compete with China, its military needs the best AI available.
Whatever the reason, the bigger question is: what happens now? The U.S. is escalating strikes against Iran, and AI is already playing a bigger role than ever. So, let's get real and talk about where exactly OpenAI’s technology could actually show up in this fight.
The Big Question: Helping to Pick Targets?
Okay, let's start with the most intense scenario: targeting and strikes.
Even with the deal in place, it’s not like you can just plug ChatGPT into a fighter jet tomorrow. The tech has to be integrated into the military’s secure, classified systems. But there’s a lot of pressure to move fast.
So, once it’s up and running, what would it actually do? I had a recent chat with a defense official that painted a pretty clear picture. Imagine an analyst with a list of potential targets. They could feed that list, along with a mountain of intelligence—text reports, satellite images, drone footage—into an OpenAI model.
Then, they could ask it to do something incredibly complex: analyze all that information and prioritize the targets. The AI could weigh logistics, like the location of specific aircraft or supplies, and spit out a ranked list of which targets to hit first.
The official was quick to add that a human would be responsible for manually checking the AI’s work. But that brings up the obvious question, right? If a person is truly double-checking every detail, how much time is this really saving? Are we just adding a high-tech rubber stamp to the process?
For years, the military has used other AI systems, like Project Maven, to do things like automatically spot potential targets in drone footage. The game-changer with models from OpenAI is the conversational part. It allows a soldier to have a dialogue with the system, asking it to interpret intelligence and recommend a course of action in plain English.
It’s hard to overstate how new this is. AI has been an analyst for the military for a while, finding needles in haystacks of data. But using generative AI to get advice on what to do in the field? That's being tested for the first time, right now.
What About Those Drones?
Another place you can bet we'll see this tech is in drone defense. Late last year, OpenAI announced a partnership with Anduril, a defense tech company that builds both drones and the systems to shoot them down.
The idea is to use OpenAI’s models to rapidly analyze data from drones attacking U.S. forces and help figure out how to stop them.
When this was announced, an OpenAI spokesperson told me it didn’t violate their policy against building systems "designed to harm others." Their reasoning? The tech is being used to target drones, not people. It’s a fine line, and one that makes a lot of people uncomfortable.
Anduril already has its own AI models trained to analyze camera feeds and sensor data to spot threats. Where OpenAI likely fits in is by providing that conversational layer—letting soldiers ask the system questions or get guidance in natural language.
The stakes here are incredibly high. In March, an Iranian drone attack that wasn't intercepted by U.S. air defenses killed six American service members in Kuwait. The pressure to improve these systems is immense.
Anduril’s central platform is called Lattice. Think of it as the operating system for modern warfare, connecting everything from drone defenses to missiles and autonomous subs. The company is landing massive contracts to layer its AI onto all sorts of military equipment. If OpenAI’s models prove useful, Lattice is designed to plug them in and roll them out across the entire battlefield, fast.
Don't Forget the Paperwork
It's not all front-line combat. A huge chunk of the military is made up of people in administrative roles—handling contracts, logistics, and purchasing. And they’re getting AI tools, too.
In December, Defense Secretary Pete Hegseth rolled out a new tool called GenAI.mil. It’s a secure way for military personnel to access commercial AI models for everyday business tasks. Think of it as a government-approved ChatGPT for the millions of people who keep the military running behind the scenes.
Google’s Gemini was one of the first models on the platform. Then came Elon Musk’s Grok. And in February, OpenAI announced its models were joining the party. They’ll be used for things like drafting policy documents, writing contracts, and assisting with mission support.
Now, someone using ChatGPT to draft a memo isn’t going to be making life-or-death decisions in Iran. But this is still a huge deal. It’s part of a relentless push from the top to embed AI into every single aspect of how the military operates. The message is clear: AI is transforming everything, from targeting decisions all the way down to the paperwork.
And as this transformation happens, OpenAI—a company once wary of military ties—is positioning itself to win a piece of it all. The debate over whether this is right or wrong is just getting started, but the technology is already on its way to the battlefield.
