Pentagon Requests $30.3M for AI-Powered Polygraph Program

The Pentagon has asked for $30.3 million across five years to build a better lie detector, according to a Department of Defense budget request. The program, referred to in budget documents as "Polygraph+" or "Polygraph Next," would pair artificial intelligence and machine learning with a technique called standoff sensing, which aims to gather physiological signals without wiring a subject to a machine. If funded, it would mark one of the US military's most direct bets yet on automated deception detection.

The money would go toward scoring algorithms rather than a single new device. Current polygraphs rely on a human examiner to interpret changes in heart rate, breathing, blood pressure and sweat, a process critics say leaves room for bias and error. The Pentagon's plan would train machine learning models to flag deception from those signals, potentially standardizing how results are read across thousands of screenings.

Standoff sensing is the more ambitious half of the request. The idea is to collect physiological data at a distance, without contact sensors or a controlled exam room, which could extend screening to settings where a traditional polygraph is impractical. Budget documents do not detail which sensors or vendors are involved, and the timeline suggests the technology is still years from fielding.

The request lands amid a long-running scientific fight over whether polygraphs work at all. A landmark 2003 National Research Council report found that polygraph evidence was too unreliable for security screening, and the technique remains inadmissible in most US courts. Federal agencies still use it widely for hiring and investigations, spending millions each year on exams that produce inconclusive results in a notable share of cases.

AI could sharpen pattern detection, but it also introduces new risks. Models trained on small, skewed datasets can inherit the flaws of the polygraph itself while adding an appearance of objectivity that juries and reviewers may overtrust. Researchers have warned that deception detection systems can perform well in lab studies and degrade sharply on real-world subjects who are anxious, medicated or trained to counter interrogation.

The budget line now moves through Congress, where lawmakers can cut, approve or redirect the funding. Watch whether the Pentagon names contractors or publishes performance benchmarks, and whether any independent validation accompanies the program before it reaches operational use. The central question remains unresolved: whether more computing power can fix a method that science has questioned for two decades.