Pentagon budget request seeks $30.3 million for Polygraph+, an AI upgrade experts call unreliable

MIT Technology Review reports, building on Inside Defense's first report, that the Department of Defense is seeking $30.3 million over five years for Polygraph+, also called Polygraph Next. If Congress approves it, the program would be run by the Defense Counterintelligence and Security Agency and would modernize federal polygraph and credibility assessment with AI and machine-learning scoring algorithms.

It would also use "standoff sensing," taking physiological readings without clipping sensors to a person. Planned uses include vetting prospective employees and detecting insider threats. The budget document has not been approved. Presage, Altec and the Defense Innovation Unit did not add to what DIU had previously disclosed about prototypes.

The request comes amid reported leak investigations. The New York Times reported that about 50 Joint Staff officers were polygraphed after coverage of U.S. weapons stockpiles in the Iran war. That context suggests demand for faster, broader screening tools.

Experts quoted by Technology Review call AI-enabled lie detection "the worst of both worlds," combining the known unreliability of polygraphs with the opacity of machine-learning systems. Polygraph results remain rarely admissible in court. The piece also notes polygraphs can be politically useful as a deterrent, regardless of accuracy.

Adding AI to an unreliable method does not necessarily make it more accurate; it can make errors harder to detect or challenge. Contactless sensing raises additional concerns, since people might be assessed without realizing it, and the program's stated insider-threat use would apply these tools to people already inside the government, not only applicants. For job applicants and employees, outcomes could affect careers based on scores they cannot easily dispute.

The request is part of the Pentagon's growing AI adoption, which spans chat assistants to screening tools, which also includes its own versions of ChatGPT and Grok and its dispute with Anthropic over usage restrictions, in which an appeals court let the Pentagon designate Anthropic a supply-chain risk. Whether Congress funds the program, what accuracy evidence is required and how results would be used in employment decisions are what to watch. More in our AI coverage and regulation coverage.