U.S. Was Unprepared for Drone Warfare, Pentagon’s Emil Michael Says

The Pentagon’s technology chief said the U.S. had been unprepared for the modern era of drone warfare and is racing to arm its troops with the autonomous systems that have come to dominate conflicts in Europe and the Middle East.

“We were caught flat-footed by the drone warfare in Ukraine and Russia,” Emil Michael, the Defense Department’s under secretary for research and engineering, said at the WSJ Tech conference Wednesday. “We weren’t prepared because we just didn’t imagine it. I just want to make sure we are imagining the next fight.”

Michael, who spoke to an audience in Napa, Calif., said the U.S. manufactured about 1,000 drones under the prior administration. Now, the country’s stockpile is up to about 500,000 drones, he said.

It’s still a small fraction of the eight million to 10 million Ukraine estimates it will build this year as the war with Russia approaches its fifth year.

“We helped the Ukrainians a ton, but they invented what they actually needed because we didn’t have it,” Michael said.

The Pentagon’s Project Meridian is tasked with studying the future of warfare, including what conflict underground or in space could look like. SpaceX Chief Executive Elon Musk and Anduril Industries founder Palmer Luckey are part of the group.

While the Pentagon is pushing into inexpensive, autonomous systems, the military is still quite conservative with its use of artificial intelligence in fighting wars, Michael said. Developing targets currently relies less on AI than intelligence reports, satellite imagery and other human-led investigatory work, he said.

“We’re not that sophisticated in doing these kinds of things,” Michael said, referring to AI-generated targets. He declined to comment on the U.S. strike in Iran last February that hit a girls’ school, which the U.S. mistook for a military target. More than 100 people were killed, including dozens of children.

The U.S. hasn’t released the findings of its investigation into the bombing. Planning for the strike had involved an AI targeting system.

“We have to have appropriate human judgment at every level. Period,” Michael said.

The Pentagon’s AI plans exclude one of the largest model developers, Anthropic, after the department earlier this year labeled the San Francisco company a national security threat. Michael reiterated his view that Anthropic is a national security risk and said it wasn’t hard to quit using the company’s models. The Defense Department still uses models from Google, OpenAI and Elon Musk’s Grok on its networks.

The U.S. currently has the lead over China in AI, Michael said, and has more data centers and better semiconductors than its top adversary. Beijing, he added, is six months to a year behind the U.S. in AI development.