As artificial intelligence continues to advance, a University of Iowa professor is leading an international conversation on how to ethically implement and regulate the technology in military applications, where it continues to have a prominent role. 

Jovana Davidovic, a UI associate professor of philosophy, published her paper investigating AI use in military technology in Nature Machine Intelligence, a scientific journal focused on research in AI, on Oct. 13. 

Davidovic’s findings come after a February AP News report found Israel has been using AI tools like OpenAI and Microsoft software to analyze surveillance and intelligence data during their military operations in Gaza and Lebanon. 

Experts in the report raised concerns that decisions surrounding targeting combatants could shift toward fully automated processes using AI, raising ethical concerns over who is responsible for potential mistakes or civilian harm.  

The UI itself is actively involved in defense-related research, including a $982,705 grant studying the detonation of energetic materials for hypersonic vehicles and a $9 million Multidisciplinary University Research Initiative on environmental films for defense‑grade materials.

Davidovic’s paper highlighted the idea of making sure soldiers have meaningful human control, a principle guaranteeing humans have the final say in when to use lethal force while using automated weapons.

“The problem has been that very often the idea of meaningful human control basically boils down to we need to make sure that the human is pushing a button right at the end,” Davidovic said. 

Davidovic said the idea of meaningful human control is unrealistic in the face of modern AI-aided warfare. Many autonomous systems are defensive, such as Navy ships, which, when detecting 300 enemy drones, would be impossible for a human to press a confirmation button to allow each shot to be fired, she said. 

Davidovic said that meaningful human control is a popular principle among policymakers who argue that fully autonomous weapons using AI violate human dignity.

 A report published in April by Human Rights Watch describes how fully autonomous AI weapons systems would degrade human dignity by delegating life and death determinants to machines.

One example Davidovic gave was loitering munitions, otherwise known as kamikaze drones. Davidovic said the drones are being used by Israel in the Israel-Hamas war and Ukraine in the Russia-Ukraine invasion.

Davidovic said an operator is able to draw a box on a screen and assign an AI-powered drone an area of responsibility where to find a target. When the drone does find a target, it will either perform a strike autonomously or ask the operator to proceed.

“Whether or not a human being in the loop increases safety is just an empirical matter,” she said. “It’s not something we can say every single time that if we have a human in the loop, it’s going to increase safety. Sometimes it will, sometimes it won’t.”

In her paper, Davidovic instead proposed the idea of appropriate human judgment, a principle she said groups like the United Nations and NATO have been moving toward over the last three to four years. 

Appropriate human judgement is a principle advocating for humans being involved at all points of an AI weapons system’s life cycle. 

From design to deployment and use, it calls for steps like engineers using discretion in building the AI and militias that purchase the systems to test the weapons adequately. 

“Appropriate human judgment is a life cycle thing,”  Davidovic said. “It’s going to require everybody in the life cycle doing their part.”

Davidovic received a $1 million grant from the Research Council of Norway to begin developing an ethical risk management framework advocating for appropriate human judgment, helping international policy makers determine the ethical risks of AI weapons. 

She began developing the framework in March and will continue until June 2028. 

“I’m hoping to put my philosopher professor hat on and my practical consulting experience and sort of provide useful guidance to defense contractors,” she said. “That will assure appropriate human judgment and therefore minimize harm to civilians.”

Davidovic said since World War II, the world has killed far too many civilians, even though weapons are advertised as being more and more precise. 

“It’s not just a matter of there’s more conflict,” she said. “It’s just that we’re not doing good on protecting civilians during wartime. Appropriate human judgment and good AI weapons show that when we do analysis at the end of the war, they perform better than humans alone on minimizing human casualties.”

Gabe Harris, the president of Applied AI, a UI student organization that helps students explore real-world AI tools and applications, said AI is a convenient tool in the everyday world, but it performs better when interacting with users rather than standing on its own two legs. 

“AI is not perfect, and it takes human specialists to actually question what it’s saying,” he said. “If we double check and fact check what AI outputs are, and continue to question them, I think that’s what encapsulates keeping humans in the loop, especially when it comes to military technology.” 

Harris said the time to begin developing ethical frameworks like that of Davidovic’s is now. 

“It’s just advancing so fast,” he said. “That fact, on its own, brings up worries when you give it military technology that can harm other humans. If we’re not confident in its decision making, then how can we be confident in giving it weapons?”

Capt. Kyle Harvey, recruiting officer and assistant professor of aerospace studies at UI, said he foresees AI being a key component in the U.S. military’s emerging weapon systems. 

“With the explosion of technology in the late 90’s and early 2000’s, we’re starting to see that now with AI,” he said. “The military has always been one of the first adopters for emerging technology, so it’s going to be one of the first adopters, in my opinion, for in-depth AI use.”

Before being assigned to Detachment 255, the UI’s Reserve Officers’ Training Corps, Harvey served as an intelligence officer.

While he was unable to comment on how AI might be integrated into general Air Force weapons systems, he sees AI integration into the intelligence field as a challenge. 

“I operate on a completely different network, being on the top secret side,” he said. “I can’t just go on ChatGPT on my system. So there’s certain career fields that don’t operate on the normal internet that wouldn’t have access to AI.”

Harvey said the data processing capabilities of AI outweigh anything currently integrated into Air Force intelligence systems. He also said large-scale AI integration could bring the risk of AI giving false information or leading to data breaches.

Steve Fleagle, associate vice president and CIO of UI’s IT services, was in the audience for one of Davidovic’s talks on AI ethics. Having heard these presentations, Fleagle called the idea of AI being integrated into large-scale weapon use frightening. 

“A good ethical position for AI is that you keep a human in the loop,” he said. “AI can make mistakes, AI can have bias, and you need a human in there to try to balance that out. So I agree 100 percent that keeping a human perspective in there should be at the center of the discussion.”

Fleagle said AI advancement is moving at a rapid pace, and the general public has shown more interest in it, underscored by over 2,000 people who have participated in HawkAI courses at the UI  since its launch in fall 2024.

Throughout this advancement, Fleagle said it is important to keep ethical considerations at the forefront of AI discussion. 

“People need to think about ethics when they’re thinking about AI, because some people just get enamored with the technology,” he said. “We’re trying to make sure that we raise awareness across the board, and make those courses approachable for anybody.”