The Pentagon wants $30 million to build an AI-powered lie detector
The Pentagon’s proposed $30.3 million investment in an AI-powered lie detector, Polygraph+, signals a renewed attempt to modernize credibility assessment technologies, but it risks repeating historical failures where unproven tools were deployed under the guise of scientific rigor. While the project promises to leverage standoff sensing and machine learning to improve accuracy, critics argue that the fundamental limitations of polygraphy—subjectivity, susceptibility to countermeasures, and lack of verifiable ground truth—remain unaddressed, raising questions about its potential misuse in high-stakes security vetting.
What is the Pentagon’s Polygraph+ project, and how does it differ from traditional polygraph tests?
Polygraph+ is a five-year, $30.3 million initiative by the Department of Defense to modernize federal polygraph and credibility assessment technologies using AI and standoff sensing, which allows physiological readings without physical contact. Unlike traditional polygraphs that rely on blood pressure, pulse, breathing, and sweat measurements attached to a subject, Polygraph+ aims to use remote sensing and machine learning to analyze multiple physiological indicators, potentially combining them into a composite deception score.
Who is leading the Polygraph+ project, and what is its intended use?
The project is being run by the Defense Counterintelligence and Security Agency (DCSA), which conducts background checks for the federal government. The DCSA plans to use the new technology for vetting prospective employees and insider threat detection, though specific technologies and implementation details remain unspecified in the budget request.
What technologies are being considered for Polygraph+, and where did they originate?
The Pentagon’s Defense Innovation Unit (DIU) previously selected two companies—Presage Technologies and Altec Research—to develop prototypes for remote deception detection. Presage Technologies uses standard cameras to measure heart and breathing rates, while Altec Research’s prototype tracks head movement, facial skin temperature, and pore activity. These technologies were chosen through an open submission process in 2023.
How reliable are current polygraph tests, and why have they been criticized?
Polygraph tests have remained largely unchanged since the 1920s and rely on physiological responses to baseline and target questions. Despite claims of 80–94% accuracy by the American Polygraph Association, studies—including a 2003 report by the US National Research Council—have found the evidence for their efficacy to be weak. Criticisms include subjectivity in interpretation, higher false-positive rates for minority groups, and the ability of interviewees to manipulate results using countermeasures. Polygraph results are also rarely admissible in court.
Ce que ça pourrait changer
The Pentagon’s push for AI-driven lie detection could normalize the use of unproven technologies in security vetting, potentially leading to widespread false accusations or the erosion of trust in institutional processes. If deployed at scale, even minor inaccuracies in such systems could have outsized consequences given the Pentagon’s 2.8 million employees. The project also reflects broader trends in surveillance and credibility assessment, where the promise of objectivity through AI may overshadow its lack of scientific validity.
