Pentagon seeks $30M to build an AI-powered lie detector
MIT Tech Review AI · rss · 2026-09-25
The US government wants to spend $30.3 million over five years on an upgraded lie detector, per a Department of Defense budget request. The program, "Polygraph+" or "Polygraph Next," focuses on AI/ML scoring algorithms and "standoff sensing" — reading physiological signals without attaching devices to a subject.
Scope
- Run by the Defense Counterintelligence and Security Agency (DCSA) for vetting applicants and "insider threat detection"; the budget has not yet been approved by Congress
- Earlier DoD efforts (2023 DIU call) picked Presage Technologies and Altec Research to prototype camera-based heart/breathing measurement and non-contact sensing of head movement, facial skin temperature and pore activity
Skepticism
- Polygraph tech has barely changed since the 1920s; a 1983 OTA report and the 2003 National Research Council found evidence for screening use "weak at best," and results are rarely admissible in court
- The American Polygraph Association claims 80–94% accuracy, but the NRC noted that at DoD scale (2.8M staff) such a test would falsely flag tens of thousands
- Interpretation is subjective, minority groups are more often judged deceptive, and trained subjects can beat the test with countermeasures
- Experts call AI plus polygraph "the worst of both worlds" — adding uncertainty on top of invalidity, with no ground truth to link physiological patterns to lying
- Researchers fear the new tools will serve as psychological props and threats rather than valid science
More from AGI Musings
- AI Agent Incidents May Usher in an Era of Less Frontier Transparency, Researchers Warn — Manderljung · 2026-09-25
- 30+ Helmholtz Munich PIs map out AI-built Virtual Human and AI-run lab workflows — Pseudomanifold · 2026-09-25
- Why smart people chasing "meta-meta-meta problems" end up solving nothing — JRIngallinera · 2026-09-25
- New book 'Perspectives of Machine Consciousness' argues LLMs may already be conscious — cccalum · 2026-09-25
- Reddit debate: neural network fundamentals never changed — is 'we don't understand AI' hype? — nezvanovova · 2026-09-25
- HN debate: Tech needs humanists more than ever — eigenBasis · 2026-09-25