Pentagon Seeks $30.3 Million for AI-Powered Polygraph+ Lie Detector
Pentagon seeks $30.3 million for AI-powered Polygraph+, despite weak evidence, bias risks and fears it could intimidate staff in internal leak probes.
Summary
On September 25, 2026, the Pentagon was seeking $30.3 million over five years for Polygraph+, or Polygraph Next, which remains subject to congressional approval. The Defense Counterintelligence and Security Agency would combine AI scoring with contactless “standoff sensing” for applicant vetting and insider threat detection, though specific technologies remain unconfirmed. Under Defense Secretary Pete Hegseth, around 50 Joint Staff officers underwent polygraphs in September after coverage of depleted US weapons stockpiles during the Iran war.
A 2023 Defense Innovation Unit competition selected Presage Technologies, whose cameras measure heart and breathing rates, and Altec Research, whose prototype tracks head movement, facial skin temperature and pore activity. Neither company, DCSA nor DIU provided further details.
Federal agencies administer tens of thousands of polygraphs annually, although results are rarely admissible in court. A 1983 congressional assessment found very limited evidence for employee screening, while the 2003 National Research Council called efficacy evidence “weak at best.” The American Polygraph Association claims 80% to 94% accuracy, but applying an imperfect system across 2.8 million DOD employees could falsely accuse tens of thousands. Examiners can disagree, minority groups face greater deception findings, and trained subjects can manipulate baselines. Earlier multimodal projects, including Silent Talker, iBorderCtrl and AVATAR, faded without reliable real-world results. Sophie van der Zee says AI could combine stress, cognitive load and concealment signals, but no universal physiological marker proves deception. Kyri Kotsoglou and Marion Oswald warn that uncertain algorithms layered onto invalid tests may primarily intimidate subjects into confessing.
Positives
- $30.3 million over five years would fund AI scoring and contactless physiological sensing through Polygraph+.
- Presage Technologies can measure heart and breathing rates using standard cameras without attaching sensors to subjects.
- Altec Research’s prototype combines head movement, facial skin temperature and pore activity measurements.
- AI could combine physiological stress, cognitive load and concealment signals into a multimodal assessment that is harder to manipulate.
- Congress must approve the Polygraph+ budget before the Defense Counterintelligence and Security Agency proceeds.
Risks & concerns
- The 2003 National Research Council found evidence for polygraph efficacy was “weak at best.”
- No universal physiological signal reliably identifies deception across all people and circumstances.
- An imperfect test applied across 2.8 million DOD employees could falsely accuse tens of thousands.
- Minority groups are more likely to be judged deceptive, while different examiners can reach sharply different conclusions.
- Trained subjects can defeat polygraphs by artificially increasing physiological responses during baseline questions.
- Polygraph+ could become an intimidation tool in leak investigations rather than a reliable source of evidence.