The Pentagon wants $30.3 million over the next five years to build an AI-powered lie detector, according to a Department of Defense budget request first detailed by Inside Defense. The program, called Polygraph+ or Polygraph Next, would pair machine-learning scoring algorithms with 'standoff sensing' — physiological readings taken without attaching any device to the subject. It would be run by the Defense Counterintelligence and Security Agency, which already conducts background checks across the federal government.
The stated purpose is to vet prospective employees and support 'insider threat detection' across a Department of Defense workforce of 2.8 million. The federal government already conducts tens of thousands of polygraph tests a year for screening, using technology that has barely changed since the device was invented in the 1920s. Congress has not yet approved the budget request, and the DCSA did not respond to a request for details on which specific technologies Polygraph+ will use.
The push comes as the Pentagon leans harder on polygraphs under Defense Secretary Pete Hegseth. The New York Times reported in September that around 50 officers on the Joint Staff had been given polygraph tests after news coverage disclosed depleted US weapons stockpiles in the war with Iran. Researchers who study deception detection are unimpressed with the plan to bolt AI onto a tool they view as scientifically shaky to begin with.
“It's a misguided effort to reduce the complex to something that is tangible.”— Kyri Kotsoglou, Professor at Northumbria Law School
Key facts
- 01The Department of Defense is requesting $30.3 million over five years for Polygraph+, a modernized lie detector using AI scoring and contactless sensors.
- 02The program would be run by the Defense Counterintelligence and Security Agency, which vets prospective employees across a workforce of 2.8 million.
- 03A 2023 Defense Innovation Unit process picked Presage Technologies and Altec Research to build prototypes tracking heart rate, skin temperature, and pore activity.
- 04A 2003 US National Research Council report called polygraph evidence 'weak at best'; humans spot lies just over half the time unaided.
- 05The New York Times reported in September that around 50 Joint Staff officers were polygraphed after leaks about depleted US weapons stockpiles in the war with Iran.
Clues about the technology stack sit in prior DoD work. In 2023, the Defense Innovation Unit ran an open call for deception-detection products and picked two prototype builders: Presage Technologies, which claims to measure heart rate and breathing from standard cameras, and Altec Research, a medical sensor firm now branching into contactless sensing. A screenshot released by the DIU shows Altec's prototype tracking head movement, facial skin temperature, and pore activity. Presage, Altec, and the DIU did not respond to requests for comment.
The underlying science remains contested. Current polygraphs measure blood pressure, pulse, breathing, and sweat, then compare responses to baseline versus target questions. In 1983, Congress's Office of Technology Assessment concluded there was very limited evidence supporting the polygraph for employee screening. A 2003 US National Research Council report called the evidence 'weak at best.' Research suggests humans can spot a lie just over half the time without any technical assistance.
The American Polygraph Association claims accuracy of 80 to 94%. Even at that level, the 2003 NRC report warned, screening at DoD scale would produce a large volume of false positives — tens of thousands of people wrongly flagged as deceptive across a 2.8-million-person workforce. Interpretations also vary sharply between examiners, and people from minority groups are more often judged deceptive. Trained subjects can also beat the test, often by heightening their physiological response to baseline questions using tricks like stepping on a pin hidden in a shoe.
“If you know how it works, you can beat it.”— Sophie van der Zee, Associate professor studying deception at Erasmus University
Sophie van der Zee, who studies deception at Erasmus University in Rotterdam, argues the polygraph's real function is deterrence — subjects often confess before the test starts. But that only holds while people believe the machine works. Decades of attempts to build better lie detectors, from thermal cameras to pupil trackers to brain scans, have not produced reliable results outside the lab. 'There is still no Pinocchio's nose,' van der Zee said.
AI could, in theory, help by finding patterns in physiological data that human examiners miss, and by combining multiple signals into a single deception score that is harder to game. Van der Zee frames lie detection as targeting three underlying things: physiological stress, cognitive load, and the conscious effort to conceal. Current polygraphs address only the first. Multi-modal AI systems could, in principle, address all three — the same pitch behind Silent Talker, a video-based system developed at Manchester Metropolitan University in the 2000s and folded into the EU's iBorderCtrl pilot, and behind AVATAR, a US border-crossing project that combined eye-tracking, voice analysis, and body movement. All of those projects quietly faded.
“Even if you have all the records in the world from polygraph tests, you don't know whether those polygraph tests are right or not.”— Marion Oswald, Professor of law
Kyri Kotsoglou of Northumbria Law School calls the AI-polygraph pairing 'the worst of both worlds,' layering algorithmic uncertainty on top of an invalid base measurement. The core problem, he and colleague Marion Oswald argue, is a lack of ground truth: even a vast dataset of polygraph results cannot train a reliable classifier if the underlying labels — lie or truth — were themselves unreliable. Oswald says the current push looks less like a scientific upgrade than an intimidation tool. She frames it as a response to the administration's concerns about leaks and loyalty, used to pressure confessions rather than surface valid information.
For AI vendors, Polygraph+ is a live federal contract in a security-adjacent category that has, until now, been dominated by a handful of legacy examiners and hardware makers. Presage and Altec are early front-runners by virtue of their DIU prototypes, and the standoff-sensing angle opens the door to computer-vision and biosignal startups that have not historically sold into the counterintelligence market. The commercial upside is real; the scientific foundation is not. If Congress approves the request, the DCSA will be spending $30 million to productize a technique that two federal science reviews, four decades apart, have already flagged as unreliable — and that a workforce of 2.8 million will now be asked to trust.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
