Pentagon Seeks $30 Million for AI-Powered Lie Detector

Summary

The US Department of Defense is seeking $30.3 million over five years for Polygraph+, also called Polygraph Next, a program intended to modernize federal polygraph and credibility-assessment technology. Run by the Defense Counterintelligence and Security Agency, it would use AI and machine-learning scoring algorithms for employee vetting and insider-threat detection, while exploring “standoff sensing” that measures physiological signals without attaching equipment to a subject. The budget request has not been approved by Congress, and the specific technologies remain unclear. Earlier Defense Innovation Unit work points to possible use of cameras and non-contact medical sensors: Presage Technologies has described measuring heart and breathing rates with standard cameras, while an Altec Research prototype tracks head movement, facial skin temperature, and pore activity. The Pentagon’s interest comes amid increased use of polygraphs to investigate alleged leaks, including tests reportedly given to about 50 Joint Staff officers. Experts quoted in the article say the underlying technology remains scientifically disputed. Polygraphs infer deception from changes in blood pressure, pulse, breathing, and sweating relative to baseline questions, but major reviews have found weak or limited evidence for employee screening. The Defense Department employs 2.8 million people, so an imperfect system could falsely accuse tens of thousands. Examiners can disagree, minority groups may be judged deceptive more often, and trained subjects can use countermeasures to manipulate physiological responses. Earlier systems combining thermal imaging, eye tracking, voice analysis, or body movement have not produced reliable results outside laboratories. AI could combine multiple signals, but researchers argue that patterns associated with stress or cognitive load cannot be reliably tied to lying without trustworthy ground truth. Critics therefore fear Polygraph+ could add algorithmic uncertainty and institutional pressure to an already unreliable method, functioning more as intimidation than as a validated source of information.