# Pentagon Seeks $30M for AI-Powered Lie Detector Upgrade DoD's Polygraph+ initiative aims to replace outdated tech with AI-driven credibility assessment tools. - **Source:** [MIT Technology Review](https://www.technologyreview.com) - **Published:** 2026-09-25 - **Topic:** [Business](https://brifing.si/c/business) - **Original article:** - **On Brifing:** https://brifing.si/s/01fc3ff2-d2d9-4f1d-ae4d-e1e4bdd0051e ## Polygraph modernization push The Pentagon's $30.3M Polygraph+ program seeks to replace 1920s-era polygraph technology with AI-powered systems using standoff sensing. This non-contact approach measures physiological responses without physical attachments. The Defense Counterintelligence and Security Agency (DCSA) will lead development for employee vetting and insider threat detection. Current polygraphs rely on obsolete metrics like blood pressure and sweat, with accuracy claims between 80-94% that researchers dispute. The upgrade responds to increased polygraph use under Defense Secretary Pete Hegseth to identify press leaks. ## Technical specifications Two prototype developers were selected through the Defense Innovation Unit: Presage Technologies analyzes heart and breathing rates via standard cameras, while Altec Research tracks head movements, facial temperature, and pore activity. The budget document reveals plans for multi-modal deception scoring combining multiple measurements. Unlike current single-angle polygraphs, the new system aims to assess physiological stress, cognitive load, and concealment efforts simultaneously. However, specific algorithms remain undisclosed, and Congress hasn't yet approved funding for the 5-year initiative. ## Historical accuracy challenges Government assessments repeatedly question polygraph reliability. A 1983 Congressional report found scant evidence for employee screening efficacy, while the 2003 National Research Council called evidence 'weak at best.' Humans detect lies 50% of the time without technology. At the Pentagon's scale (2.8M employees), even 90% accuracy could falsely accuse thousands. Minority groups face higher false-positive rates, and trained subjects can manipulate results using countermeasures like concealed pins. The American Polygraph Association's accuracy claims remain scientifically contested. ## Academic skepticism Legal scholar Kyri Kotsoglou calls AI-enhanced polygraphs 'the worst of both worlds,' layering algorithmic uncertainty atop invalid methodology. Erasmus University's Sophie van der Zee notes polygraphs primarily function as psychological deterrents rather than scientific tools. Without ground truth data, machine learning can't reliably correlate physiological patterns with deception. Previous attempts like the EU's iBorderCtrl and US AVATAR systems—which incorporated eye tracking and voice analysis—failed to achieve reliable real-world results despite initial promise. ## Current polygraph limitations Traditional polygraphs measure responses to baseline questions ('Is the sky blue?') versus target questions ('Did you commit a crime?'), but results vary wildly between examiners. The tests' inadmissibility in US courts reflects scientific doubts. Recent use cases include screening 50 Joint Staff officers after reports about depleted Iran war stockpiles. Professor Marion Oswald warns new systems may replicate polygraphs' misuse as 'psychological props' to extract confessions rather than generate valid intelligence—especially concerning given the administration's focus on leaks and perceived disloyalty. ## Implementation risks The DCSA's planned deployment for employee vetting raises concerns about algorithmic bias and due process. Past systems like Silent Talker showed how deception scoring can encode examiner subjectivity into software. Without transparency about training data or error rates, AI could amplify existing disparities. The program's budget justification cites 'modernization' needs but doesn't address fundamental validity questions. Legal experts warn such tools might disproportionately impact minority applicants and create new vectors for false accusations in security clearance processes. ## Industry participation Presage Technologies and Altec Research represent emerging players in non-contact biometrics seeking government contracts. Their prototypes illustrate the shift from direct physical measurement to computer vision-based analysis. However, neither company responded to inquiries about their technologies' validation processes. The Defense Innovation Unit's selection criteria remain undisclosed, raising questions about how commercial products will be evaluated against mission requirements. Previous defense AI projects have faced challenges transitioning from lab demonstrations to operational systems. ## Broader implications The Polygraph+ initiative reflects growing defense sector interest in behavioral analytics, but replicates patterns of prior failed projects. Without addressing core scientific limitations—like the absence of universal deception indicators—AI risks automating flawed methodologies. The program's focus on insider threats may prioritize security over employee rights, particularly given polygraphs' history of coercive use. As van der Zee notes, effective deception detection requires triangulating multiple physiological and cognitive signals—a challenge no current technology has solved outside controlled environments. ## Sources covering this story - https://www.technologyreview.com/2026/09/25/1145144/pentagon-ai-lie-detector/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._