In 2015, a group of prominent AI and robot-ics researchers signed an open letter warn-ing of the dangers of autonomous weapons. “The key question for humanity today,” they wrote, “is whether to start a global AI arms race or to prevent it from starting. If any major mil-itary power pushes ahead with AI weapon devel-opment, a global arms race is virtually inevitable.”1 Today, many nations are working to apply AI for military advantage, and the term “AI arms race” has become a catchphrase used by both critics and proponents of AI militarization. In 2018, then-Un-der Secretary of Defense Michael Griffin, calling for the United States to invest more in AI, stated, “There might be an artificial intelligence arms race, but we’re not yet in it.”2 In a 2020 Wired article, Will Roper, then chief acquisition officer for the U.S. Air Force, warned of the risks of falling behind in a “digital arms race with China.”

The so-called AI arms race has become a com-mon feature in news headlines,4 but the arms race framing fails to match reality. While nations are clearly competing to develop and adopt AI tech-nology for military use, the character of that com-petition does not meet the traditional definition of an arms race. Military AI competition nevertheless does pose risks. The widespread adoption of mili-tary AI could cause warfare to evolve in a manner that leads to less human control and to warfare be-coming faster, more violent, and more challenging in terms of being able to manage escalation and bring a war to an end. Additionally, perceptions of a “race” to field AI systems before competitors do could cause nations to cut corners on testing, lead-ing to the deployment of unsafe AI systems that are at risk of accidents that could cause unintend-ed escalation or destruction. Even if fears of an “AI arms race” are overblown, military AI competition brings real risks to which nations should attend. There are concrete steps nations can take to miti-gate some of these dangers.