Pentagon reportedly wants $30 million to build an AI-powered lie detector · Digg

Pentagon reportedly wants $30 million to build an AI-powered lie detector

MIT Technology Review says Polygraph+ will use new sensor technologies and AI to vet staff and hunt leakers.

1 Source, 1h ago, first seen 1h ago

A prototype interface tracks head movement, facial temperature and pore activity for remote polygraph sensing.

TLDR

The Pentagon wants $30 million to build Polygraph+, an AI-powered lie detector, MIT Technology Review reports. The planned system would combine new sensor technologies and AI to vet staff and hunt leakers. Experts cited by the outlet call AI-enabled lie detection “the worst of both worlds.”

The U.S. Department of Defense is seeking $30.3 million over five years for Polygraph+, a program that would combine machine learning with new sensors in an attempt to improve employee screening and leak investigations. Congress has not yet approved the requested funding.

MIT Technology Review reports that the Defense Counterintelligence and Security Agency would run the project. The budget documents describe scoring algorithms and “standoff sensing,” which means measuring physiological signals without attaching traditional polygraph equipment to a subject.

New sensors do not solve the old validity problem

Earlier Defense Innovation Unit prototypes used ordinary cameras to estimate heart and breathing rates or tracked signals such as facial temperature, pore activity and head movement. Polygraph+ could combine several measurements into one deception score.

The scientific objection is more fundamental than sensor quality. Traditional polygraphs measure stress-related signals, not lies, and researchers do not have dependable ground-truth labels showing whether the people in large screening datasets were actually deceptive. Training a model on examiners' judgments can therefore automate existing assumptions without validating them.

The federal government conducts tens of thousands of polygraph screenings each year, even though major scientific reviews have found weak evidence for their reliability in employee screening. Researchers interviewed by MIT Technology Review warned that an AI layer could make an uncertain process appear more objective while preserving false positives, examiner bias and the possibility of countermeasures.

Today's Rank

—

Not ranked yet

Combined views

3.2K

1 Source, first seen 1h ago

2 likes

Pentagon reportedly wants $30 million to build an AI-powered lie detector

MIT Technology Review says Polygraph+ will use new sensor technologies and AI to vet staff and hunt leakers.

1 Source, 1h ago, first seen 1h ago

TLDR

New sensors do not solve the old validity problem

1 Source

@techreviewPolygraph+ will use new sensor technologies and AI to vet staff and hunt leakers. Experts say AI-enabled lie detection is “the worst of both worlds”. https://trib.al/ApW7TmB

Today's Rank

—

Not ranked yet

Combined views

3.2K

1 Source, first seen 1h ago

2 likes

Pentagon reportedly wants $30 million to build an AI-powered lie detector

TLDR

New sensors do not solve the old validity problem

Combined views

Pentagon reportedly wants $30 million to build an AI-powered lie detector

TLDR

New sensors do not solve the old validity problem

1 Source

Combined views

New sensors do not solve the old validity problem

New sensors do not solve the old validity problem

Sentiment

Summary

Sentiment

Summary

Sentiment

Summary

1 Source

Sentiment

Summary