The Pentagon wants to spend $30.3 million over the next five years on a new generation of lie detection technology, betting that artificial intelligence and machine learning can do what decades of conventional polygraph testing have struggled to achieve: reliably distinguish truth from deception.
According to a Department of Defense budget request, the program is called "Polygraph+" or "Polygraph Next." It would focus on scoring algorithms powered by AI and machine learning, alongside a technique known as "standoff sensing," which can gather physiological readings from a person without requiring them to be physically attached to a device. In theory, that could make lie detection faster, less intrusive and more adaptable to modern security screening. In practice, it places the Pentagon squarely in a field with a long history of technical promise and persistent scientific skepticism.
The stated goal is straightforward enough. The Defense Department wants to improve the accuracy and reliability of polygraph assessments, which remain a staple of sensitive government vetting even though their scientific foundations have long been contested. Polygraphs do not directly detect lies; they measure physiological responses such as heart rate, breathing and skin conductance, then infer whether those responses suggest deception. That distinction has fueled criticism for years, because stress, fear, confusion or even the pressure of being questioned can produce the same signals.
The new proposal suggests the Pentagon is not abandoning that model so much as trying to modernize it. By adding AI-based scoring, officials hope to reduce human inconsistency and improve the interpretation of results. By exploring standoff sensing, they may also be trying to move beyond the traditional setup in which a subject must be wired to sensors. The broader ambition is clear: to make lie detection more scalable, more discreet and perhaps more persuasive to agencies that rely on it.
But the project also arrives with a warning label written by history. Technology has repeatedly promised to solve the problem of deception detection, and repeatedly fallen short. From early polygraph machines to more recent attempts using voice analysis, facial cues or brain imaging, the central challenge has remained the same: human behavior is messy, and there is no universally accepted biomarker for lying. AI may improve pattern recognition, but it does not automatically solve the underlying scientific problem of whether those patterns actually map to deception.
That tension is especially important in a national security context, where false positives can have serious consequences. If a system wrongly flags an innocent person, it can damage careers, delay clearances or shape investigations in ways that are hard to reverse. If it misses deception, it can create a false sense of confidence. The Pentagon's interest in a more advanced system suggests that existing tools are not satisfying its needs, but it does not guarantee that the next version will be any more definitive.
The budget request reflects a broader pattern across government and industry: a belief that AI can improve judgment in domains where humans are inconsistent, biased or overwhelmed. Yet lie detection is one of the most difficult of those domains, because the target is not a stable physical object or a clear visual pattern, but an internal mental state that people can mask, mimic or misread. That makes the Pentagon's proposal as much a test of the limits of AI as a test of deception itself.
For now, Polygraph+ remains a request, not a breakthrough. But it underscores how deeply the desire for a machine that can tell truth from lies continues to shape public spending, even as the evidence for such a machine remains elusive. The Pentagon may be hoping that AI can finally sharpen an old tool. The larger question is whether the tool can ever become something it has never truly been: a reliable detector of lies.