The Pentagon wants $30 million to build an AI-powered lie detector

THE NEWS

What happened

The US Department of Defense has requested $30.3 million over five years to develop Polygraph+, an AI-enhanced lie detection system designed for vetting employees and detecting insider threats without physical contact. Despite historical skepticism regarding polygraph reliability and the lack of a definitive biological marker for deception, the Pentagon aims to use machine learning to analyze physiological stress, cognitive load, and concealment efforts. Critics warn this initiative risks amplifying existing biases and using unproven technology as a psychological tool to intimidate staff rather than accurately identifying truth.

CONTEXT

Why it matters

The US military is investing $30.3 million over five years into Polygraph+, an AI-driven system designed to detect lies without physical contact. The program targets insider threats and employee vetting under the Defense Counterintelligence and Security Agency. While proponents hope machine learning can find hidden patterns in physiological data, experts caution that polygraphs lack a definitive biological marker for lying. With current technology struggling to distinguish truth from stress reliably, critics fear this initiative may serve more as an intimidation tactic than a valid investigative tool, potentially exacerbating biases against minority groups.

AT A GLANCE

Key facts

• The Pentagon requested $30.3 million for the Polygraph+ program over the next five years.
• The system aims to use 'standoff sensing' to take physiological readings without attaching devices to subjects.
• Defense Secretary Pete Hegseth has increased reliance on polygraphs to investigate alleged leaks, including those regarding US weapons stockpiles.
• Previous attempts at non-contact lie detection, such as the 2023 DIU prototypes by Presage Technologies and Altec Research, have not yielded reliable results outside controlled labs.
• Experts note that current polygraph technology has remained largely unchanged since the 1920s and its results are rarely admissible in court.
• Research indicates humans can detect lies slightly better than chance, but AI lacks a 'ground truth' to validate its patterns against lying behavior.
• Critics argue combining AI with polygraphs creates 'the worst of both worlds,' adding uncertainty to an already flawed technology.