The U.S. Department of Defense has requested $30.3 million over the next five years to develop a next-generation lie detection program known as Polygraph+ or Polygraph Next. According to budget documents, the project aims to integrate artificial intelligence and machine learning into credibility assessments, alongside “standoff sensing” technology, which allows for physiological monitoring without the need for physical attachments to the subject.

The Defense Counterintelligence and Security Agency (DCSA) is slated to oversee the program, intending to utilize the technology for employee vetting and the identification of insider threats. This push for modernization occurs amid heightened internal scrutiny regarding information leaks, with reports indicating that the Pentagon has increased its use of traditional polygraph testing for staff members.

While the Pentagon has not disclosed specific technical details, previous efforts by the Defense Innovation Unit (DIU) provide insight into the potential direction of the project. In 2023, the DIU selected companies such as Presage Technologies and Altec Research to develop prototypes capable of tracking physiological markers like heart rate, breathing, facial temperature, and pore activity through standard cameras.

Advertisement

The initiative faces skepticism from experts who question the scientific validity of automated deception detection. Kyri Kotsoglou, a legal scholar at Northumbria University, described the integration of AI into polygraph testing as “the worst of both worlds,” arguing that it compounds existing issues of invalidity with the uncertainty of algorithmic processing. Similarly, law professor Marion Oswald suggested that such tools often function more as psychological instruments for intimidation than as reliable scientific devices.

Critics also point to the historical failure of similar projects, such as the EU-funded iBorderCtrl and the U.S.-based AVATAR, which attempted to use multi-modal analysis to detect deception but ultimately saw limited success. Despite these concerns, proponents suggest that combining AI with multi-modal data—such as physiological stress, cognitive load, and behavioral cues—could theoretically offer a more comprehensive approach than traditional methods. However, as researcher Sophie van der Zee noted, there remains no universal physiological indicator for lying, often referred to as a “Pinocchio’s nose.”

Source: MIT Technology Review