Pentagon Seeks $30M for AI Lie Detector

Overview

The U.S. Department of Defense has submitted a budget request seeking $30.3 million over the next five years to develop an advanced form of lie detection technology. The program, tentatively referred to as "Polygraph+" or "Polygraph Next," aims to modernize federal credibility assessment tools by integrating artificial intelligence and machine learning scoring algorithms.

A significant portion of the initiative focuses on standoff sensing, a technique that allows examiners to capture physiological readings from a subject without attaching physical sensors to their body. This shift represents a move away from the invasive procedures typical of traditional polygraph tests, which have remained largely unchanged since their invention in the 1920s.

The Technology Behind Polygraph+

According to details of the budget document, first reported by Inside Defense, the project seeks to improve the accuracy and reliability of federal polygraphs. The new system will be administered by the Defense Counterintelligence and Security Agency (DCSA), which currently conducts background checks for the federal government.

While the specific technologies have not been finalized, previous efforts by the Pentagon’s Defense Innovation Unit (DIU) offer insight into what this next generation of deception detection might look like. In 2023, the DIU selected two companies to build prototypes capable of non-contact sensing:

• Presage Technologies: This company claims to measure heart rate and breathing rate using standard video cameras, leveraging computer vision rather than direct contact.
• Altec Research: A medical sensor company branching into non-contact technologies, Altec’s prototype tracks head movement, facial skin temperature, and pore activity.

This evolution addresses one of the major limitations of current polygraphs. Traditional devices rely on exogenous sensors to measure blood pressure, pulse, respiration, and sweat. These metrics are then interpreted by an examiner who compares responses to baseline questions against target questions. AI-driven multi-modal detection aims to combine these various data streams into a single, more robust deception score that is harder for subjects to manipulate.

"The more you can have combined methods that approach it from these three different angles—physiological stress, cognitive load, and conscious effort to conceal—the more successful the detection is likely to be," says Sophie van der Zee, an associate professor studying deception at Erasmus University.

Political Context and Insider Threats

The push for enhanced lie detection comes during a period of heightened internal tension within the Department of Defense. Under Defense Secretary Pete Hegseth, the Pentagon has increasingly relied on polygraph tests to identify sources of alleged leaks to the press.

In September, The New York Times reported that approximately 50 officers on the Joint Staff were subjected to polygraph examinations following news coverage regarding depleted U.S. weapons stockpiles in the conflict with Iran. The new AI-enhanced tools are expected to be used primarily for vetting prospective employees and detecting insider threats, rather than for judicial proceedings.

However, critics argue that the administration is placing undue faith in tools with questionable scientific validity. Kyri Kotsoglou, a professor at Northumbria Law School, described the effort as "misguided," noting that it attempts to reduce the complexity of human deception into something tangible and measurable.

Scientific Skepticism and Historical Failures

Despite technological advancements, the fundamental problem of lie detection remains unsolved. There is no single physiological marker of lying that applies universally; researchers often refer to this as the lack of a "Pinocchio’s nose." While the American Polygraph Association claims an accuracy rate between 80% and 94%, the 2003 report by the U.S. National Research Council (NRC) concluded that evidence supporting the polygraph’s efficacy is "weak at best."

The stakes of inaccurate testing are particularly high within the military context. With 2.8 million employees, even a small error rate in a screening system could result in tens of thousands of false accusations. Furthermore, polygraph interpretation is inherently subjective, with different examiners often arriving at wildly different conclusions from the same data.

Studies also indicate that individuals from minority groups are disproportionately judged as deceptive under current protocols. Additionally, trained subjects can employ countermeasures, such as artificially elevating their physiological responses to baseline questions by stepping on a hidden pin, thereby confusing the resulting graphs.

"If you know how it works, you can beat it," van der Zee explained. She noted that the primary benefit of polygraphs is often deterrence rather than detection; subjects frequently confess before the test even begins because they believe the machine will reveal their guilt.

Historical attempts to replace or enhance the polygraph—ranging from thermal cameras to brain scans—have largely failed to produce reliable results outside of controlled laboratory settings. Human lie detection skills, even without technical assistance, hover just above 50% accuracy, suggesting that the cognitive complexity of deception remains beyond the grasp of current technology.

As Congress reviews the $30.3 million request, the debate continues over whether AI can finally unlock the secrets of truthfulness or if it will merely automate an inherently flawed process.