The U.S. Department of Defense has asked Congress for $30.3 million over the next five years to fund “Polygraph+” – an AI‑enhanced lie‑detector that would modernize the agency’s polygraph and credibility‑assessment tools. The request, detailed in a recent budget document, aims to add machine‑learning scoring algorithms and a “standoff sensing” capability that could capture physiological data without attaching sensors to a subject.
What happened- The Pentagon’s budget request earmarks $30.3 million for the Polygraph+ program, also called “Polygraph Next.”
- The project will be managed by the Defense Counterintelligence and Security Agency (DCSA), which conducts background checks for the federal government.
- Funding would support AI‑based scoring algorithms and non‑contact sensing methods that measure heart rate, breathing, facial temperature, pore activity and head movement.
- The effort follows a 2023 Defense Innovation Unit (DIU) open‑submission that selected Presage Technologies and Altec Research to build prototypes using standard cameras and medical‑sensor tech.
Why it mattersPolygraph testing, a practice dating back to the 1920s, relies on physiological cues such as blood pressure, pulse, breathing and sweat. While the American Polygraph Association claims high accuracy rates, multiple government reviews—most notably a 1983 Office of Technology Assessment report and a 2003 National Research Council assessment—have found the evidence supporting polygraph efficacy to be weak. At the scale of the DoD’s 2.8 million staff, even a modest error rate could result in tens of thousands of false accusations.
Critics highlight several systemic problems:
- Subjectivity: Different examiners can reach wildly different conclusions from the same data.
- Bias: Minority groups are statistically more likely to be judged deceptive.
- Countermeasures: Trained subjects can deliberately alter physiological responses, such as by applying a hidden pin to raise baseline stress.
- Legal standing: Polygraph results are rarely admissible in court.
Adding AI does not automatically resolve these issues. Researchers note that AI can detect patterns invisible to human examiners, but without reliable ground‑truth data—i.e., certainty about when a subject is truly lying—machine‑learning models may simply amplify existing uncertainties. Professors Kyri Kotsoglou and Marion Oswald warn that AI‑enhanced polygraphs could become “the worst of both worlds,” serving more as intimidation tools than scientific instruments.
The bigger pictureThe Pentagon’s renewed focus on polygraph testing coincides with heightened internal security concerns. Under Defense Secretary Pete Hegseth, the department has increasingly turned to polygraphs to investigate alleged leaks, such as the recent New York Times report that around 50 Joint Staff officers underwent testing after coverage of depleted U.S. weapons stockpiles in the Iran war.
Earlier DoD initiatives provide context for Polygraph+:
- The 2023 DIU competition sought commercial solutions for deception detection, selecting companies that demonstrated non‑contact sensing via cameras and medical‑grade sensors.
- Past research projects—including the UK’s “Silent Talker,” the EU’s iBorderCtrl pilot, and the U.S. AVATAR program—explored multi‑modal deception scoring by combining video, eye‑tracking, voice analysis and body movement. These efforts have largely faded, underscoring the technical challenges of reliable lie detection.
Proponents argue that AI could enable “multi‑modal” approaches, integrating physiological stress, cognitive load and conscious concealment cues into a single deception score. Yet, as scholars like Sophie van der Zee emphasize, no single physiological marker universally signals lying, and the technology’s deterrent effect relies largely on the belief that it works—not on proven accuracy.
What happens nextThe budget request has not yet been approved by Congress, and the DCSA has declined to comment on the specifics of the planned technology. If funded, Polygraph+ would proceed alongside ongoing DIU collaborations with Presage Technologies and Altec Research, whose prototypes already demonstrate non‑contact measurement of heart rate, breathing, facial temperature and pore activity.
Future steps may include:
- Development and testing of AI‑driven scoring algorithms on data collected via standoff sensing.
- Integration of the system into DCSA’s background‑check and insider‑threat workflows.
- Potential expansion of testing to additional DoD personnel if initial results are deemed satisfactory.
However, scholars caution that without robust validation, the program could repeat the historical pattern of deploying controversial lie‑detection tools that serve more as psychological leverage than reliable evidence. The outcome will likely hinge on whether the AI components can demonstrably improve accuracy beyond the longstanding limitations of traditional polygraphs.
