The US Department of Defense has requested $30.3 million over the next five years to develop an AI-enhanced polygraph system called Polygraph+ or Polygraph Next, according to the agency's budget submission. Rather than replacing traditional polygraph technology entirely, the program focuses on developing advanced scoring algorithms that apply machine learning to physiological data collection and interpretation. The system would analyze heart rate variability, skin conductance levels, and respiratory patterns with greater granularity than conventional polygraphs, using AI models trained on historical examination data to improve detection accuracy. The DoD framed the investment as a solution to a mounting backlog in security clearance investigations, with an estimated 700,000 pending cases as of 2023—a bottleneck that has delayed deployments and slowed hiring across federal agencies.
However, polygraph researchers and civil liberties advocates have raised substantial concerns about the approach. Dr. Drew Richardson, a former polygraph examiner and researcher at the National Academy of Sciences, has publicly stated that polygraphs achieve only 60-70 percent accuracy under ideal conditions, and 'adding AI to a fundamentally unreliable tool doesn't resolve its core problems—it automates bias.' The American Civil Liberties Union and the Center for Democracy and Technology have warned that AI scoring algorithms trained on existing examination data may perpetuate historical disparities, as research suggests polygraph outcomes are influenced by factors including examiner bias, cultural background, and anxiety disorders rather than deception alone. These groups argue that automated decision-making in security clearance contexts raises due process questions and could unfairly disqualify qualified applicants.
The Pentagon's investment reflects a broader pattern of federal agencies adopting AI for high-stakes screening processes, with mixed results. The Amazon hiring tool scandal of 2018, in which an automated recruiting system systematically discriminated against women, and subsequent controversies over algorithmic bias in military target identification systems have set precedent for congressional scrutiny. The DoD has faced similar criticism over its JIED (Joint Intelligence Enterprise Defense) and algorithmic warfare initiatives. Critically, Pentagon officials have not publicly disclosed whether Polygraph+ has undergone independent validation testing against ground truth data, or whether the system has been piloted with known-innocent and known-guilty populations under controlled conditions. Without such evidence, the $30 million commitment appears to rest primarily on the assumption that machine learning can solve a problem that may be inherent to the underlying methodology itself.