Pentagon Requests $30M for AI-Powered Polygraph+ Lie Detector
Department of Defense seeks machine learning and standoff sensors to modernize federal credibility assessment and curb leaks.
The United States Department of Defense has requested $30.3 million over the next five years to develop "Polygraph+"—a modernized credibility assessment system powered by artificial intelligence and non-contact physiological sensors. Aimed at vetting prospective staff and detecting insider threats across 2.8 million personnel, the program represents a high-stakes effort to replace century-old lie detector technology with machine learning scoring algorithms. The initiative comes amid heightened security tensions within the Pentagon following widespread leak investigations and aggressive polygraph sweeps ordered by military leadership.
Key Details
According to federal budget request documents, the Polygraph+ program will be managed by the Defense Counterintelligence and Security Agency (DCSA), which oversees background checks and security clearances across federal agencies. Rather than relying on traditional physical contact sensors like blood pressure cuffs and breathing straps, Polygraph+ will emphasize "standoff sensing"—techniques capable of capturing physiological signals remotely without touching the subject.
Key facts surrounding the $30.3 million Pentagon initiative include:
- Funding Allocation: A $30.3 million request spanning five fiscal years to modernize credibility assessment technology.
- Operational Scope: Application across employee vetting, security clearance renewals, and insider threat monitoring for 2.8 million Department of Defense military and civilian personnel.
- Prototype History: Previous Defense Innovation Unit (DIU) contracts engaged computer vision firm Presage Technologies to monitor respiration via standard video cameras and Altec Research to track facial skin temperature, head movement, and pore activity.
- Context of Urgency: Increased reliance on polygraphs by Defense Secretary Pete Hegseth, including recent mandatory testing for Joint Staff members investigating leaked reports on US weapons stockpiles.
What This Means
Traditional polygraph tests have remained fundamentally unchanged since their invention in the 1920s. Despite widespread use across federal intelligence and defense agencies, scientific consensus—including major evaluations by the National Research Council—has long categorized polygraph efficacy as weak and subjective. By introducing machine learning models, defense officials hope to reduce human examiner bias and detect subtle physiological patterns undetectable by eye.
However, legal and scientific experts caution that combining machine learning with polygraphy risks creating "the worst of both worlds." Because traditional lie detection lacks a ground-truth physiological signal for deception, training machine learning models on legacy polygraph datasets threatens to encode existing subjective errors into automated scoring models. Rather than producing objective truth, critics argue the system may function primarily as an algorithmic psychological deterrent.
Technical Breakdown
Polygraph+ seeks to overcome the limitations of single-metric polygraphs by building multi-modal AI models that process multiple non-contact telemetry streams simultaneously:
- Optical Physiological Telemetry: Utilizing standard high-resolution cameras and computer vision algorithms to estimate heart rate variability, micro-expression shifts, and breathing patterns remotely.
- Thermal and Pore Tracking: Deploying specialized infrared sensors to monitor minute skin temperature changes and localized sweat gland activity on the face.
- Multi-Modal Data Fusion: Integrating physiological stress, cognitive load indicators, and behavioral concealment markers into a unified AI deception score.
- Countermeasure Resistance: Algorithmic detection designed to identify deliberate breathing manipulation or physical countermeasures used by subjects to manipulate baseline scoring.
Industry Impact
The launch of Polygraph+ signals a broader trend toward non-contact, AI-driven biometric surveillance in government and high-security enterprise environments. For defense contractors and commercial AI developers, the initiative opens a lucrative market for computer vision, thermal analysis, and multi-modal sensor fusion platforms.
Simultaneously, the program raises serious concerns for defense personnel and government technologists. Applying probabilistic AI models to employee vetting at the scale of 2.8 million workers inevitably introduces false-positive risks. In an imperfect system, even a minor margin of error could falsely flag thousands of innocent staff members as security risks, leading to revoked clearances, stalled careers, and legal challenges regarding algorithmic due process.
Looking Ahead
As Congress considers the Department of Defense budget request, Polygraph+ will face scrutiny from privacy advocates, legal scholars, and civil liberties groups. Lawmakers are expected to press the Defense Counterintelligence and Security Agency on validation standards, false-positive rates, and whether AI-assisted lie detection can meet legal evidentiary thresholds.
Whether Polygraph+ delivers a genuine scientific breakthrough or simply digitizes historic polygraph flaws, its development marks a pivotal moment in the deployment of artificial intelligence for institutional security. As defense agencies rush to safeguard classified information in an era of rapid technological change, the boundary between automated threat detection and algorithmic intimidation will remain a central point of contention.
Source: MIT Technology Review(opens in a new tab) Published on ShtefAI blog by Shtef ⚡
