Pentagon Wants $30 Million for AI-Powered Lie Detector "Polygraph+"

The US Department of Defense wants to spend $30.3 million over the next five years on an AI-enhanced lie detector, according to a DoD budget request that has not yet been approved by Congress. The program, called "Polygraph+" or "Polygraph Next," will focus on scoring algorithms that use artificial intelligence and machine learning, and on "standoff sensing" — taking physiological readings from a subject without attaching a device to their body.

Details of the budget document, first reported by Inside Defense, state that the project will "modernize federal polygraph and credibility assessment technologies" to improve their accuracy and reliability. The Defense Counterintelligence and Security Agency (DCSA), which conducts background checks for the federal government, will run the program. According to the budget document, the technology will be used for vetting prospective employees and "insider threat detection." The DCSA did not respond to a request for more information, and it is not yet clear which specific technologies will be deployed.

The stakes are considerable. The Department of Defense employs 2.8 million staff, and any screening tool deployed at that scale will produce errors in bulk — a concern that has shadowed polygraph technology for four decades. Critics say the new program repeats a pattern of failed attempts to technologize lie detection.

"It's a misguided effort to reduce the complex to something that is tangible," says Kyri Kotsoglou, a professor at Northumbria Law School in the UK who studies the use of polygraphs in the justice system.

A program born amid leak hunts

The budget request lands at a moment of high tension inside the Pentagon. Under Defense Secretary Pete Hegseth, the department has increasingly turned to polygraph tests to find the sources of alleged leaks to the press. In September, the New York Times reported that around 50 officers on the Joint Staff had been given polygraph tests after news coverage revealed depleted US weapons stockpiles in the war with Iran.

Marion Oswald, a professor of law who has written with Kotsoglou on polygraph use in the justice system, sees a direct connection between the political climate and the new funding push. "It seems very much a response to the concern of the current administration to leaks and perceived lack of loyalty," she says. "[Lie detection is] being used as a threat, to intimidate and force people to confess to things, as opposed to anything that's actually getting valid information."

Clues from the Defense Innovation Unit

Other DoD efforts offer hints about what Polygraph+ might look like. In 2023, the Defense Innovation Unit (DIU) ran an open submission process to find companies with products usable for deception detection. It selected two companies to build prototypes: Presage Technologies, which claims it can measure heart rate and breathing rate using standard cameras, and Altec Research, a medical sensor company now moving into non-contact sensing.

A screenshot of Altec's prototype released by the DIU shows the system tracks head movement, facial skin temperature, and pore activity. Presage Technologies and Altec Research did not respond to requests for comment. The DIU declined to comment.

That prototype points to the standoff-sensing ambition at the heart of Polygraph+: readings taken at a distance, without wires or cuffs, scored by algorithm.

A century-old device with a weak scientific record

Current polygraph technology has barely changed since the device was invented in the 1920s. Examiners measure blood pressure, pulse, breathing, and sweat, then judge whether a subject is lying based on differences in physiological response to baseline questions like "Is the sky blue?" and target questions like "Have you ever committed a crime?"

The federal government runs tens of thousands of these tests a year while screening employees, but the polygraph has been repeatedly debunked, and its results are rarely admissible in court. In 1983, Congress's Office of Technology Assessment concluded there was very limited evidence supporting the polygraph's use for employee screening. In 2003, the US National Research Council said evidence of its efficacy was "weak at best."

The accuracy numbers tell their own story. Research suggests humans can spot a lie just over half the time with no technical assistance. The American Polygraph Association claims the polygraph is between 80 and 94% accurate. But the 2003 NRC report noted that a screening test at that accuracy level still produces large numbers of mistakes — applied across 2.8 million DoD employees, an imperfect system could falsely accuse tens of thousands of people.

The problems compound from there. Polygraph interpretation is often subjective: different examiners get wildly different results from the same data, and people from minority groups are more likely to be judged deceptive. The test is also beatable. With training, interviewees can learn countermeasures — most commonly artificially heightening their body's physiological response to baseline questions, for example by stepping on a pin hidden in their shoe.

"If you know how it works, you can beat it," says Sophie van der Zee, an associate professor who studies deception at Erasmus University in Rotterdam. She says the machine's biggest effect is deterrence — often, subjects confess before the test even begins. "But that only works if people think a polygraph works."

Why AI, and why it might not help

Decades of attempts to build new lie detectors — thermal cameras, pupil trackers, brain scans — have yielded no reliable results outside the lab. The obstacle is structural: there is no single telltale sign of lying that holds true for everyone, all the time. "There is still no Pinocchio's nose," says van der Zee.

AI could theoretically improve polygraphs if it found patterns in the data that human examiners cannot. Algorithms are also better suited to "multi-modal" deception detection, which combines multiple measurements into an overall deception score that is harder to game. Van der Zee identifies three signals that lie detection tries to capture: physiological stress, cognitive load, and the conscious effort people make to conceal lying. Current polygraph technology measures only the first.

"The more you can have combined methods that approach it from these three different angles, the more successful you will be," she says.

The multi-modal idea is not new. In the 2000s, researchers at Manchester Metropolitan University developed "Silent Talker," a system that generated a deception score from video footage; it was later folded into iBorderCtrl, an EU-funded pilot. In the US, a project called AVATAR combined eye-tracking, voice analysis, and body movement detection for border crossings. All of these projects have quietly faded away.

Kotsoglou argues that combining AI with the polygraph produces "the worst of both worlds," adding uncertainty on top of invalidity. Even if machine learning finds previously unseen patterns in physiological data, it cannot reliably link those patterns to lying because of a lack of ground truth.

"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," says Oswald. She fears new forms of lie detection will follow the polygraph's path and serve as a psychological prop rather than a scientific tool.

With Congress yet to approve the request, Polygraph+ still needs funding. If it proceeds, the program will test whether machine learning can succeed where four decades of deception-detection research — and a century of polygraph practice — have not.