Geoffrey Hinton says Ukraine made military AI “more complicated”

Geoffrey Hinton, the computer scientist often called the “godfather of AI,” says Russia’s war in Ukraine changed his stance on military AI. In an interview with NBC News published Tuesday, Hinton said the relationship between AI and the military is “more complicated” than he used to think.

That is a big signal from someone who has spent years warning against military applications of the technology, including pushing for an international ban on lethal autonomous weapons. Hinton told NBC News: “I used to think that we should try very hard to prevent lethal autonomous weapons, but if you look at what’s happening in Ukraine, it becomes much more complicated.” The key detail is not theoretical. He said Ukraine is surviving because of drones, and that if modern warfare is about drones, “it’s very hard to argue that one country should refuse to do it.”

To understand why this matters beyond the ethics debate, you have to look at how Ukraine’s defense actually works, at least as Hinton described it. He acknowledged the value of AI-driven systems on the battlefield, but he also pointed to the reality that Ukraine’s defense relies on the large and constant production of munitions, including drones that are largely piloted and directed by humans. In other words, Hinton is not saying “autonomous killing machines solved everything.” He is saying the battlefield is forcing tradeoffs, and those tradeoffs are messy.

That phrase, “a mess,” is doing a lot of work. Hinton said he is still uneasy about AI’s role in modern warfare, even if drones and AI-enabled capabilities have become pivotal in Ukraine’s defense. The uncomfortable second-order effect of that admission is that it reframes what “responsible” AI in defense even means. If the technology helps a country survive against an invading force, then a blanket refusal starts to look like a self-imposed disadvantage, not a moral stance. But if the systems also reduce human direction and decision-making, they can also enable killing more people much faster, which is exactly the kind of risk Hinton has been warning about.

That tension is exactly why the conversation around AI and defense keeps splitting into camps. On one side are people focused on preventing harm and slowing down lethal autonomous weapons, including Hinton’s previous calls for an international ban. On the other side are governments and defense organizations arguing that AI can help protect warfighters and improve precision. That is not just theoretical policy talk. The source notes that over the past decade, employees at tech companies including Google and Microsoft pushed back against defense-related projects, including some involving AI, which shows how difficult it has been for the industry to align engineering priorities with public safety and military use.

Now, the regulatory and policy backdrop is tightening. Last week, President Donald Trump signed a directive aimed at accelerating the development and use of AI for national security purposes. The AI, the National Security Presidential Memorandum reads: “will be among the most transformative technologies to national security in the history of the United States.” It continues: “When adopted appropriately, AI can help protect our warfighters during peacetime and on the battlefield, enable precise operations that minimize harm to civilians, and ensure the United States continues to maintain technical overmatch against our adversaries and strategic competitors.”

Even if you personally agree or disagree with that framing, the direction is clear: national security is becoming a primary growth engine for AI deployment. That has already changed how companies may interpret risk. The source highlights two examples. First, the Pentagon labeled the US company Anthropic a supply chain risk after it requested that its Claude AI not be used on autonomous weapons. Second, it gutted an office meant to reduce harm to civilians during war. Those moves suggest that the “how” of compliance, not just the “whether,” is getting more intense. If a lab tries to restrict use cases, regulators and the defense bureaucracy may respond by treating the company as less aligned with national security goals.

So what does Hinton want now? He is still pushing engagement and education, saying the public pressure is what will rein in the big AI companies. He told NBC News: “The only thing that’s going to rein in those big AI companies is public pressure,” adding, “I see my mission as educating the public, so they understand the dangers of AI, as well as understanding the good things.” That is a pragmatic shift. Ukraine does not just provide a case study for policy. It provides a proof point that AI-enabled drones and battlefield technology can move outcomes quickly, and it makes moral arguments feel less like abstract frameworks and more like urgent operational decisions.

For executives and boards, Hinton’s update is a reminder that the risk conversation is no longer one-dimensional. It is now shaped by geopolitical incentives, procurement pressures, and public scrutiny, all moving at the same time. If your company touches AI systems that could be repurposed for defense, procurement, or national security programs, then the question is not whether the military will use AI. The question is how quickly the political and technical environment will normalize those deployments, and what your governance, product boundaries, and public messaging look like when they are tested in real time.