Pentagon leaders are racing to insert artificial intelligence into nearly every corner of the military, even as lawmakers scramble to figure out guardrails.
Defense Department officials insist any slowdown in AI development could cede control to adversaries, dismissing calls from industry leaders and members of Congress to pause and better understand the long-term ramifications of the computer systems.
That tension is playing out in real time as the Pentagon faces increasingly bold threats from China and Russia. But the military’s rush to stay ahead brings separate risks if it fails to safeguard hastily implemented AI tech.
“Winning and staying ahead in the AI race is probably one of the most important things for the United States to do, both from a military and economic security perspective,” Air Force Secretary Troy Meink said. “That is a race that the U.S. can’t lose.”
But critics warn the rush to use advanced AI in national security operations could also open sensitive military systems to new threats, especially if leaders aren’t deliberate about installing defenses during upgrades.
“The potential for out-of-control rogue agents invading websites or degrading systems applies to the Pentagon as much as any other organization,” said Sen. Richard Blumenthal (D-Conn.), who sits on the Senate Armed Services Committee. “The Pentagon could be a very positive force for risk evaluation. But so far, it has avoided that responsibility.”
Congress is considering guardrails for the military’s AI use as part of annual defense policy legislation, expected to become law this year. Those restrictions would include a new review process to ensure that any autonomous systems and AI capabilities have “human judgment” at the center of all decisions, as well as new security standards and guidance for any AI agents embedded in military systems.
“We’re going to be seeing a lot more coming in the future,” said Sen. Mike Rounds (R-S.D.), who chairs the cybersecurity panel for the Senate Armed Services Committee. He called those proposals a positive first step for regulating military AI.
But the provisions in the authorization bill likely won’t become law until December, a slow timeline for many lawmakers concerned about the rapidly evolving systems. House Minority Leader Hakeem Jeffries this week called for Congress to “immediately” pass new regulations for AI development in the wake of fears that the technology could end humanity if allowed to advance unchecked.
Defense Department officials — along with President Donald Trump — have pushed back on those dire warnings, vowing that humans remain in control of the advanced tools.
“Everybody who is assuming that not using AI is the safest option is really picking the most dangerous course of action we can take because our adversaries are moving in this direction,” Cameron Stanley, the Pentagon’s chief digital and AI officer, said at a recent conference in Washington. “We’ve been very clear: We’re going to be an AI-first force…That means taking on some risk, but it’s not taking on ridiculous risks.”
The Pentagon did not respond to a request for follow-up comment.
European allies — whose own security planning is tied to American AI advances — have voiced similar support for that approach. The increasing strain on traditional munitions stockpiles across Europe and America, a result of the Iran war, has fueled desires for more high-tech military options.
“I do not see a possibility of a slowdown,” said Giedrimas Jeglinskas, a former Lithuanian lawmaker and NATO official. “The technology is too important for economic growth, competitiveness and national security.”
Industry leaders have adopted that philosophy as well, even as they urge caution. Anthropic’s policy chief Sarah Heck said at POLITICO’s Decoded Summit on Wednesday that the U.S. “can’t do safety from second place” and must continue to press ahead with development of new models.
But some at the summit acknowledged the seeming contradiction. Rep. Greg Casar (D-Texas), House sponsor of the Ban Artificial Superintelligence Act, said lawmakers’ goal should be “a world where there are not autonomous killer robots anywhere” instead of ensuring that Americans hold a military edge on that kind of technology.
“The stakes are high here,” he said. “We should be doing what is right.”
Audrey Decker and Jack Detsch contributed to this report.