US Senators Seek Probe Into Military AI Failures

Senators seek probe of US military AI failures

Washington, Sep 22

Senators Mark Warner, Jack Reed and Chris Coons asked Defence Secretary Pete Hegseth and Director of National Intelligence Jay Clayton to provide inspectors general unrestricted access to examine the reported incidents.

Warner is the senior Democrat on the Senate Intelligence Committee. Reed is the ranking member of the Senate Armed Services Committee, while Coons leads Democrats on the Senate Appropriations defence subcommittee.

Their letter cited public reports that an AI targeting platform relied on outdated geographical data during a US kinetic strike in Minab, Iran.

It also referred to an aborted military interdiction operation after allegedly false information generated by AI was included in US intelligence distributed by Special Operations Command Pacific.

The senators did not provide further operational details about the Minab strike or the aborted interdiction mission in their statement. The reported failures have not been independently established in the material released by them.

"Recent events - including public reports of an AI-targeting platform relying on outdated geospatial data in a kinetic strike in Minab, Iran and a public report on an aborted military interdiction operation driven by potential AI-hallucination in disseminated intelligence by U.S. Special Operations Command Pacific - have given us growing concern about the extent to which agencies under your oversight have prioritized acceleration of AI capability adoption and 'experimentation' over effective governance," the senators wrote.

They warned that errors in targeting and intelligence systems could cause operational failures and weaken trust in American intelligence among the US public and international partners.

"Public reports that AI targeting platforms have repeatedly generated spurious outputs (based on outdated input data or outright hallucination) risk serious operational error," they wrote.

Such failures could also undermine confidence "in the accuracy and reliability of U.S. intelligence and intelligence-informed kinetic actions," the senators said.

They called for immediate investigations by the relevant inspectors general and greater public disclosure of the eventual findings.

The lawmakers also referred to what they described as Pentagon pressure on a leading American AI company that had resisted allowing its technology to be used in fully autonomous lethal operations. The statement did not identify the company.

"These potential mistakes - against the backdrop of the Department of Defense's harassment of a leading U.S. AI vendor over the latter's reluctance to allow its products' use in fully-autonomous kinetic operations - deserve immediate investigation by relevant Inspectors General," the senators wrote.

They said AI could play an important role in intelligence analysis but must remain subject to rigorous professional standards and human oversight.

"While AI has appropriate applications in the analytic and other fields, strong tradecraft standards should ensure that no disseminated intelligence product contains AI-generated errors, particularly when such intelligence may serve as the impetus for a kinetic action against a nuclear-armed adversary," they wrote.

The senators criticised the administration's revocation of National Security Memorandum 25 and a Defence Department strategy that, they said, treats governance requirements as obstacles to faster AI adoption.

"A 'Move Fast and Break Things' approach may serve for high-engagement memes on social media platforms, but it risks potentially grave operational miscalculations," they wrote.

They asked investigators to examine the two reported incidents and determine whether other significant failures involving AI-enabled targeting had occurred without being publicly disclosed.

The US military has increasingly used artificial intelligence to process surveillance information, identify patterns and support operational decisions. The expansion has intensified debate over data accuracy, accountability and the degree of human control required when AI-generated intelligence could contribute to the use of lethal force.

The issue also carries implications for American defence partners, including India, as militaries expand cooperation in intelligence, surveillance and emerging technologies. Confidence in shared systems will depend partly on whether AI-generated assessments can be independently checked before they influence military action.

— IANS