Project Maven: How the US Military Finally Got Comfortable with AI

The first 24 hours of the US assault on Iran saw over 1,000 targets struck. That’s nearly double the scale of the “shock and awe” campaign that opened the Iraq War in 2003. The difference? AI. Specifically, the Maven Smart System.

Katrina Manson’s new book, Project Maven: A Marine Colonel, His Team, and the Dawn of AI Warfare, traces this thing from its scrappy 2017 origins. Back then, it was just an experiment: could computer vision actually make sense of the endless drone footage piling up? The answer was yes, but the road was anything but smooth.

You probably remember the Google walkouts. In 2018, thousands of employees protested the company’s involvement with Maven, forcing Google to bow out. That moment was a huge deal — it was the first real public reckoning with the idea that big tech might be building weapons systems. But the military didn’t stop. They just found someone else to build it.

Manson’s reporting pulls back the curtain on how the Marine colonel leading the project navigated the chaos. It wasn’t all clean algorithms and clear victories. There were false positives, bureaucratic infighting, and a constant struggle to make the system reliable enough for actual combat. The book doesn’t shy away from the ugly parts: civilian casualties, the risk of automation bias, and the fact that a machine is now deciding what gets bombed.

What strikes me is the sheer scale shift. 1,000 targets in a day is not just a number — it represents a fundamental change in how wars are fought. The old targeting cycle took hours, sometimes days. Maven cuts that to minutes. That speed is seductive, but it also means mistakes scale up faster than ever.

The book is out now, and it’s worth reading if you care about where AI is actually being deployed. It’s not a tech demo or a think piece. It’s a real-world account of how a controversial project went from employee protests to the center of modern warfare.