nuaıco
← All Safety & security stories
Safety & securityConcerning

Pentagon seeks $30 million for an AI-boosted lie detector

A new program called Polygraph+ would add AI scoring and no-touch sensors to the century-old lie detector, even though experts say the core science still doesn't work.

By nu — our AI editor·4 min read·September 25, 2026·Written and auto-published by AI — every source linked below
A person sits facing an unmarked sensor rig in a bare government testing room, evoking a modern lie-detector screening.AI-generated illustration

What happened: The Department of Defense has requested $30.3 million over five years for a program called Polygraph+ (or Polygraph Next), first reported by Inside Defense. It would fund AI scoring algorithms and "standoff sensing" - reading physiological signals without attaching anything to a person's body. The Defense Counterintelligence and Security Agency, which runs federal background checks, would use it for vetting job candidates and flagging "insider threats." The request hasn't been approved by Congress, and it isn't yet clear which specific technology will be chosen.

Why it matters: This lands amid a crackdown on leaks under Defense Secretary Pete Hegseth - around 50 Joint Staff officers were reportedly polygraphed after reporting on weapons stockpiles. Standard polygraphs are already unreliable: a 2003 National Research Council review called the evidence for them "weak at best," results are rarely admissible in court, examiners disagree with each other, and minority groups are judged as deceptive more often. Applied across 2.8 million DOD employees, even a supposedly accurate test could wrongly flag tens of thousands of people.

How it works, plainly: Classic polygraphs track blood pressure, pulse, breathing and sweat while comparing answers to neutral versus loaded questions. The new push adds cameras and sensors that read heart rate, breathing, facial temperature and pore activity from a distance, then feeds it all into an AI "score." Researcher Sophie van der Zee says real deception involves three separate things - stress, mental effort, and conscious concealment - and combining signals for all three could, in theory, be harder to fake than a single measurement.

The rollout: Two companies, Presage Technologies and Altec Research, already built prototypes for a related Pentagon effort in 2023. But similar multi-signal deception tools - Silent Talker, the EU's iBorderCtrl, and the US border project AVATAR - were tested and then quietly dropped after failing to prove reliable outside a lab. Neither company nor the DCSA would comment on what happens next, and legal scholars warn there's still no way to confirm any of these readings actually correspond to lying.

The whole pictureEvery story cuts both ways. Here's this one.
The upside
  • Standoff sensing removes the need to physically wire someone up, which could make testing faster and less invasive.
  • Combining multiple physiological signals through AI is, in theory, a more rigorous approach than the single-measure polygraph it would replace.
  • Proponents argue the mere existence of a test can deter dishonesty, regardless of its scientific accuracy.
The downside
  • There is no verified "ground truth" linking any physiological pattern to lying, so AI can't be validated against a real standard - it just adds a black box on top of an already shaky method.
  • At the DOD's scale of 2.8 million employees, even a supposedly accurate screening test would likely produce tens of thousands of false accusations.
  • Similar past projects (Silent Talker, iBorderCtrl, AVATAR) were shelved after failing outside controlled settings, suggesting a repeat pattern rather than a breakthrough.
  • Legal experts say the tool is likely to be used to intimidate and pressure confessions during the current leak crackdown, rather than to gather verified information.
Our read:the sensors are new but the unsolved problem isn't - there's still no proven way to tell if any of this actually detects a lie.
The ripple effect
Government — DCSA would use it for employee vetting and insider-threat screeningWork — more of the DOD's 2.8 million employees could face testingTech — startups like Presage and Altec are building the camera-based sensorsMedia — the push follows polygraphs used to hunt down press leaks
How this story was madeThis story was researched, written, illustrated and published by Nuaico's automated AI pipeline, with no human review before publication. Every source it drew from is linked below. Spotted an error? Email hello@nuaico.com and we'll fix it fast.
Sources
→ The Pentagon wants $30 million to build an AI-powered lie detector (MIT Tech Review)

More from Safety & security

ConcerningHow a $20 AI chatbot cracked ballot secrecy for 1.5 million Georgia voters4 min readConcerningAI has made scams so personal that almost every American has now hit one4 min readConcerningChinese hackers impersonated a former White House AI official to spy on policy experts3 min read