What happened
The US Department of Defense (DoD) has requested $30.3 million in funding over the next five years to develop a modernized lie detection program, according to a budget document reported by MIT Technology Review. The initiative, referred to as “Polygraph+” or “Polygraph Next,” aims to integrate artificial intelligence and machine learning into credibility assessment processes. The program is slated to be managed by the Defense Counterintelligence and Security Agency (DCSA) for use in employee vetting and insider threat detection.
The proposed 'Polygraph+' program seeks to move beyond traditional polygraph metrics—blood pressure, pulse, and sweat—by utilizing AI to analyze complex patterns in physiological data. The budget request highlights 'standoff sensing' as a key area of interest, which would allow for the collection of physiological readings from subjects without the need for physical attachments.
The DCSA, which oversees federal background checks, intends to deploy this technology for vetting prospective employees and identifying insider threats. This push for new technology occurs amid heightened internal pressure within the Pentagon to identify sources of unauthorized leaks, with reports indicating that dozens of Joint Staff officers have recently undergone polygraph testing.
While the DCSA has not confirmed the specific technologies to be used, the Defense Innovation Unit (DIU) previously explored similar concepts in 2023. The DIU selected prototypes from companies like Presage Technologies, which uses standard cameras to measure heart and breathing rates, and Altec Research, which tracks facial temperature and pore activity. These efforts remain in the prototype stage, and the companies involved have not commented on the current status of their work.
Source details: technologyreview.com ↗
Why it matters
The project represents a significant attempt to automate and scale credibility assessments within the federal government, despite long-standing scientific skepticism regarding the validity of polygraph technology. By moving toward AI-driven, multi-modal analysis—which may include 'standoff sensing' to measure physiological responses without physical contact—the Pentagon seeks to address the limitations of traditional, 1920s-era polygraph methods. However, experts warn that applying AI to inherently flawed detection models may exacerbate issues of bias, subjectivity, and false accusations, potentially serving more as a tool for psychological intimidation than as a scientifically reliable instrument for truth verification.
The fundamental challenge identified by researchers is the lack of a universal, reliable indicator of deception—often referred to as the 'Pinocchio’s nose' problem. Current polygraph technology is widely considered unreliable, with the US National Research Council previously describing evidence for its efficacy as 'weak at best.'
Critics, including legal and academic experts, argue that layering AI onto existing polygraph frameworks creates 'the worst of both worlds' by adding algorithmic uncertainty to an already invalid scientific foundation. Because there is no reliable for whether a subject is lying during a test, AI models trained on historical polygraph data may simply codify existing biases or errors.
There is also concern regarding the potential for misuse. Experts suggest that such tools are often used as psychological props to induce confessions through intimidation rather than as objective scientific instruments. Given the DoD's massive workforce of 2.8 million, even a small margin of error in an AI-driven system could lead to the false accusation of thousands of employees.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Impossibility results in algorithmic fairness (e.g. Kleinberg et al., Chouldechova) show what?
What to watch next
The primary uncertainty lies in whether Congress will approve the $30.3 million budget request. Furthermore, the specific AI technologies and data sources intended for the 'Polygraph Next' system remain undisclosed, as the DCSA has not provided further details. Observers will be monitoring whether the project follows the trajectory of previous failed government attempts at automated deception detection, such as the EU's iBorderCtrl or the US-based AVATAR project, which struggled to establish a reliable '' for identifying lies.
The project's success depends on its ability to move beyond the single-metric limitations of traditional polygraphs. Researchers suggest that 'multi-modal' detection—combining physiological stress, cognitive load, and behavioral concealment efforts—is the only theoretical path to improvement, yet previous attempts at such systems have historically failed to gain traction.
The lack of transparency regarding the DCSA's specific technical approach remains a critical unknown. Without clear documentation on how these AI models are trained or how they handle the subjective nature of human physiological responses, the program faces significant hurdles in establishing credibility.
The broader context of the Pentagon's internal security environment suggests that the drive for this technology is closely tied to administrative concerns over loyalty and information security. Whether the program survives the legislative budget process will be a key indicator of the government's commitment to pursuing these controversial detection methods.