The Pentagon wants to spend $30.3 million over the next five years on a new generation of lie-detection technology that would blend artificial intelligence, machine learning and remote sensing into a system the Department of Defense calls Polygraph+ or Polygraph Next. The request, disclosed in a budget document, points to a quiet but consequential push inside the U.S. national security establishment to modernize one of the most contested tools in government screening and intelligence work.
AI Credibility Push
The program is designed to improve on conventional polygraph testing by developing scoring algorithms that can analyze physiological and behavioral signals with the aid of AI. It also seeks to advance a technique known as standoff sensing, which aims to detect cues from a distance rather than through direct contact with a subject. In practical terms, that could mean a system that tries to infer deception from remotely gathered signals, rather than relying solely on the wires, sensors and examiner judgment associated with traditional polygraphs.
The budget request is notable not only for the technology it proposes, but for the institutional ambition behind it. Polygraph testing has long occupied an uneasy place in U.S. security policy: widely used in some government settings, yet frequently criticized by scientists for its imperfect accuracy and by civil liberties advocates for its potential to produce false positives and false confidence. By attaching AI and machine learning to the concept, the Pentagon appears to be betting that data-driven scoring can reduce human inconsistency and improve reliability.
That is a difficult promise to make. Machine learning systems are only as strong as the data used to train them, and deception is not a cleanly measurable phenomenon. Stress, fear, cultural differences, medical conditions and examiner bias can all affect outcomes. If the new system is meant to be used in sensitive screening environments, those limitations could become even more consequential, especially if the technology is treated as an authoritative signal rather than one input among many.
Standoff Sensing Ambitions
The inclusion of standoff sensing suggests the Pentagon is also interested in reducing the intrusiveness of current methods. Remote detection technologies have long attracted defense and intelligence funding because they promise faster, less invasive assessments in operational settings. But they also raise a familiar problem: the farther a system moves from direct measurement, the more it must infer intent from indirect indicators, and the more room there is for error.
The budget line for Polygraph Next is relatively modest in Pentagon terms, but the implications are broader than the dollar figure suggests. A five-year research and development effort gives the department time to explore whether AI can meaningfully improve credibility assessment, and whether a hybrid of algorithmic scoring and remote sensing can survive scientific scrutiny. It also reflects a broader defense trend: the search for machine-learning tools that can compress human judgment into repeatable, scalable outputs.
That trend has accelerated across the U.S. government as agencies look for AI systems that can sort information, flag anomalies and support decision-making in high-stakes environments. Yet lie detection remains one of the most fraught use cases for automation. Unlike image recognition or logistics forecasting, it deals with a deeply human question that resists simple pattern matching. The Pentagon's proposal therefore sits at the intersection of technical optimism and longstanding skepticism.
High Stakes, Low Certainty
For now, the request is a budget proposal, not a deployed capability. Congress will determine whether the program is funded, and the eventual shape of Polygraph+ will depend on how the Defense Department defines success. If the aim is merely to improve examiner support tools, the project may be easier to justify. If it is intended to produce a more definitive machine-based lie detector, the scientific and ethical hurdles become much steeper.
The broader significance lies in what the request reveals about the government's appetite for AI in sensitive domains. The Pentagon is not just asking for smarter software; it is asking whether machine learning can help solve a problem that has eluded reliable resolution for decades. That makes Polygraph Next less a finished system than a test case for the limits of frontier AI in national security.