AI’s dominance on battlefields from Ukraine to the Middle East is fueling fears that the future of war is more dangerous due to reduced human oversight.
As the Pentagon rushes to introduce physical AI, Democratic senators are trying to limit legislative challenges, recently introducing two bills they hope will curb its use in the military. Arms control experts and advocates praised the efforts, but some warned the RS that the workarounds included in the bill could weaken the law or render it ineffective.
Lawmakers challenge regulations
Last week, Sen. Kirsten Gillibrand (New York) Safe and Responsible Military AI Act,it is Limit the use of AI nuclear weapons launches, domestic surveillance, autonomous weapons systems, etc. Semi-autonomous systems, non-lethal systems, or systems used for so-called defensive purposes are not restricted.
“The most important decisions that impact national security and the lives of military personnel must always be made by humans, not unaccountable machines.” Gillibrand said In a press release. “We must act now. Rather than stifle technological progress, we must put humans in charge and establish clear rules to keep the use of AI in warfare wise and safe.”
Michael Clare, senior visiting fellow at the Arms Control Association, said the bill is “ambitious in that it sets out a broad framework for the military’s future use of AI.” To do so, he said, “we will identify categories of AI applications that have ‘significant consequences’,” such as command and control of nuclear weapons, lethal targeting, and cyber operations.
If the military wants to use AI for these “severe consequence” actions, Gillibrand’s bill would will be needed Senior Pentagon leaders must first approve the decision and notify Congress. Specifically, if the Department of Defense uses or deploys fully autonomous weapons systems, the Secretary of Defense would need to petition Congress to pass a joint resolution authorizing it.
Nuclear weapons policy experts praised the mechanism.
“Given the high-consequence nature of all nuclear weapons issues, it is appropriate for Congress to hold the Department of Defense accountable for and justifying the adoption of new AI tools in these areas,” Xiaodong Liang, senior policy analyst at the Arms Control Association, told RS.
Roberto Gonzalez, an anthropology professor at San Jose State University and author of War Virtually, a book about how technology and automation will transform warfare, was more skeptical.
“Why would an Under Secretary of Defense like Emile Michael — yes, a former Uber executive — pass legislation regulating military AI when he has the power to approve AI uses with ‘severe consequences’ such as lethal target support and nuclear command and control?” Gonzalez asked.
Michael, former Silicon Valley executive in close relationship to defense contractors, have insisted firmly Use of AI in the military field.
“It is misleading to suggest that the proposed bill would ‘prohibit’ the military from using AI,” Gonzalez said.
The Hill reported last week that Gillibrand: intend He plans to propose parts of the bill as an amendment to the next National Defense Authorization Act (NDAA), a mandatory bill that sets national defense priorities for next year.
Sen. Elissa Slotkin (D-Mich.) introduces her own AI Guardrails Act back in marchwhich is also prevent The Pentagon prohibits the use of autonomous weapons to kill “without an appropriate level of human judgment and oversight.” That too Forbidden Prevents AI from being used to deploy nuclear weapons or spy on “individuals or groups within the United States” without lawful reason.
slotkin told NOTUS Presumably, the language of the law will be included in its basic text. NDAA.
“The base text is [bipartisan] We agreed,” Slotkin said. said Notus. “If it had been controversial, it wouldn’t have been in this article. So I took this as a real sign that people knew we needed limits on the left and right, reasonable limits that weren’t overly authoritarian.”
Slotkin’s bill would also allow the Secretary of Defense to: issue a waiver It limits the use of autonomous weapons systems. The secretary must: demonstrate That’s what it takes to claim there are “extraordinary circumstances” that affect U.S. national security.
Observers were divided on the merits of the Slotkin bill.
“Suggests ‘an appropriate level of human judgment and supervision.'” [in military contexts] “It’s problematic that legislation is mandated without clearly defining what ‘appropriate’ means,” Gonzalez told RS. “This is a loophole that will likely render the proposed legislation meaningless.”
“The exceptions in both bills are very strict,” Clare told RS in written comments. “While we do allow for exceptions under extraordinary circumstances, such exceptions will require credible explanation from senior Department of Defense officials.”
But Clare conceded that “the Pentagon will always find a way around Congress if it has the determination to do so.”
John Laming Chappell, a consultant at the Civilian Center for Conflict, told RS that national security bills often contain exemptions.
“However, when the stakes are high, as in the case of deploying military applications of AI that have the potential to kill civilians with little human intervention, it is preferable to require Congress to vote affirmatively to authorize the use of the system, rather than giving the president latitude with exemptions limited only to presidential decisions and reporting requirements,” he wrote.
Enforcement is another concern.
“We should also ask what sanctions and penalties will be imposed on military personnel who violate these laws,” González said. “If a penalty is nothing more than a reprimand or a slap on the wrist, what’s the point?”
William Hartung, a senior fellow at the Quincy Institute, which specializes in the arms industry, told RS that the bill was “flawed. [they’ve] There was a loophole. ”
“At least they’re putting something out there that can be discussed and improved,” he concluded. “I give them credit.”
America’s battle over military AI intensifies
Legislation is increasingly being promoted public skepticism About AI in general and its use in military contexts. There are widespread concerns that the technology could hand over human judgment on deadly wartime decisions to less sensible people. often inaccurate machine.
The Trump administration does not appear to share these concerns. AI has been used extensively in the US-Israel war against Iran, with the US using AI-powered targeting tools to Attack over 1,000 targets the first 24 hours of the conflict. The United States is also using AI, make a plan for military operations The kidnapping of Venezuelan President Nicolas Maduro in January.
President Donald Trump signed an executive order on AI earlier this week, declared This technology should be “quickly deployed to counter any threat” to the US: “We refuse to suppress it.” [AI] “Innovation with Overly Burdensome Regulations,” Executive Order said.
Major US AI companies are also caught up in this problem.
Back in March, the Department of Defense cut off Anthropic, a prominent AI company with government contracts the request Its AI technology must not be used for fully autonomous military targets or espionage against Americans. The Pentagon insisted that its technology be turned over to AI companies that work with the Pentagon. used For “all lawful purposes” — a provision the Federation of American Scientists says vague.
Department of Defense’s move toward blackball Anthropic — the Department of Defense Supply chain risk — There is a long and ongoing legal battle between the parties.
Anthropic’s AI industry competitors (Google, OpenAI, Microsoft, SpaceX, etc.) I have it all The Department of Defense agreed that AI technology can be deployed on sensitive networks for any lawful purpose.
