Three of the Pentagon tasks surveyed by GAO included weapons for counter-camouflage (higher picture) and goal recognition (decrease photographs).

Government Accountability Office/GAO

disguise caption

toggle caption
i

Government Accountability Office/GAO

The Pentagon is constructing weapons and tasks with synthetic intelligence capabilities that officers say do not always intently adhere to the company’s own tips for AI, according to a congressional watchdog report printed in October.

AI researchers and worldwide organizations have lengthy advocated for strict regulation of AI when used in weapons that could become unpredictable or operate with little human management, arguing that these instruments could endanger the lives of troopers and civilians and threat rapidly escalating wars. But Defense Secretary Pete Hegseth has repeatedly dismissed earlier warning of AI instruments and inspired fast autonomous innovation at the Pentagon.

The Pentagon issued a memo in January saying the division would become “an ‘AI-first’ warfighting force across all components, from front to back.” Last month, Hegseth announced a new warfare command targeted on autonomous weapon improvement, including drones.

Secretary of Defense Pete Hegseth arrives on stage for a dialogue during an annual Army assembly on October 14, 2026 in Washington, DC.

Kevin Dietsch/Getty Images North America

disguise caption

toggle caption
i

Kevin Dietsch/Getty Images North America

The watchdog report, printed by the U.S. Government Accountability Office on Oct. 14, grants a uncommon look inside how these varieties of tasks are being developed.

The GAO surveyed 10 officers concerned with constructing 10 weapons and instruments for the Pentagon that use AI to mechanically determine targets on the battlefield, course of surveillance and enhance tactical resolution making, and other duties. The tasks are among more than 300 being constructed with AI in the Pentagon.

When responding to the GAO survey, six of the 10 officers said their groups had “identified risk(s)” with utilizing the tasks “in an operational environment.”

To tackle some of these types of dangers, in 2021, the Pentagon despatched a memo outlining six primary tenets to guarantee that weapons constructed with AI are created responsibly. Those six tips primarily stress the significance of Pentagon-wide oversight, verification and “real-time monitoring” of the know-how, and cooperation with allies “to advance global norms grounded in shared values.”

Alissa Czyz, who directed the GAO evaluate, said the tenets developed from earlier moral rules, and while they weren’t guidelines, they were understood to be essential by individuals concerned with constructing AI instruments at the Pentagon.

“It’s a memo, right, at the end of the day, so it’s not a department instruction,” said Czyz. “It’s something that the department has signaled that should be implemented.”

Most of the individuals concerned with making the 10 weapons and tasks reviewed by GAO had taken “at least initial steps” to observe the six tenets outlined in the memo, according to the report, and some of the tasks were in early phases, Czyz clarified.

But some of the responses indicated officers weren’t always intently following the tips.

Four officers instructed the GAO that necessities for the weapons and tasks they were working on weren’t “aligned with Responsible AI tenets.” And three of the 10 officers instructed the GAO they “don’t know” if the tasks will be constantly monitored after they’re deployed, though the six tenets encourage monitoring.

Officials instructed the GAO that more entry to experience about the company’s tips or workers who know about them would improve their means to incorporate AI responsibly. But the group that helps handle schooling on the Pentagon’s tips, the Department of the Responsible AI Division, is understaffed. Although the group was approved to rent 10 individuals in 2023, as of February, only three of these positions were crammed and the Pentagon “is unlikely to fill vacant positions or increase staffing,” the report states.

“That’s a very small number of folks tasked with a large responsibility of responsible AI throughout the department,” said Czyz.

The GAO issued two suggestions: The Pentagon should more clearly define the duties and assembly schedules of the workplace charged with overseeing AI regulation, and tackle understaffing. Although the Pentagon agreed with these suggestions, as of Oct. 22, the company had not carried out these modifications, a GAO consultant confirmed to NPR.

NPR requested the Pentagon when it deliberate to make these modifications, but did not obtain a response.

When delivering a speech at SpaceX in January, Hegseth indicated he believed AI improvement at the Pentagon should not be slowed by these types of steps.

“Our enterprise currently operates on staffing and committee cycles measured in months and years, and that’s unacceptable,” said Hegseth. “Military AI is going to be a race for the foreseeable future where the risks to U.S. national security of moving too slowly outweigh the impacts of imperfect alignment.”

NPR would like to hear from more individuals who work with federal companies or AI corporations and have info to share about how companies are utilizing AI. You can e mail the reporter of this article, Chiara Eisner, at ceisner@npr.org or contact her on Signal here, username ceis.78.