Pentagon denies using AI lie-detection in polygraph revamp

Overview

The Pentagon has clarified that certain artificial intelligence tools previously described as part of its polygraph modernization efforts are not included in the current program overhaul, following inquiries from Defense News about earlier statements and declassified presentation slides.

As of July, the Department of Defense had been preparing AI tools designed to analyze text, voices, and faces to assess staff trustworthiness, according to declassified slides and prior DoD statements.

DCSA Statement

On Wednesday, the Pentagon's Defense Counterintelligence and Security Agency (DCSA)—which vets security clearance holders and applicants—issued an emailed statement clarifying the scope of the "Modernizing Polygraph" effort:

• Does not incorporate generative super intelligence (artificial intelligence) that creates new content such as emails, graphics, or voiceovers
• Does not include large language models (LLMs), such as ChatGPT or similar chatbots trained on massive datasets
• Does not involve facial action coding of emotions, micro-expression analysis, or vocal analyses

The statement added that while "those technologies are not part of the effort," DoD will "continue to monitor emerging scientific capabilities to evaluate their potential utility down the road."

DCSA did not address other types of AI models that, according to the slides, researchers had been developing to scan spoken and written language for "deceptive speech" patterns.

Background and Context

Wednesday's statement came after Defense News sought responses to criticisms from national security lawyers, scientists, and AI auditors regarding AI's ability to analyze deception—defined as not merely lying, but doing so knowingly and willfully.

Critics warned that AI-aided scanning of speech patterns, vocal stress, or facial expressions to read minds:

• Lacks scientific backing
• Will compound errors
• May violate civil liberties in a domain where few such rights exist

The recent DCSA statement contrasts with several earlier DoD statements, FY2027 budget documents, and DCSA slides obtained by Defense News through an open records request.

Prior Pentagon Statements

In July, a Pentagon official speaking on condition of anonymity described the advantages of AI-driven sentiment and deception analysis:

"AI processes data in real time and enables standardized, objective data analysis. AI-based analysis is a force multiplier for our human investigators, not a replacement."

The official added that "the models are primarily trained on data from laboratory-controlled studies using volunteer participants."

DCSA slides, including presentations from 2021 through 2025, show that the Air Force, DCSA, and academic labs had been working on "sentiment analysis" AI models and "deceptive speech" AI-aided analysis since 2019, as part of a "Credibility Assessment Modernization" project.

In July, the DoD official described the AI models as part of a larger multi-year "transformation," referred to interchangeably as "Polygraph+" and "Credibility Assessment Modernization."

Budget and Technology

Current budget materials price the Polygraph+/Polygraph Next project at $31 million, including costs for:

• AI "scoring algorithms" and "decision aids"
• Thermal imaging tools to detect blood flow changes reflecting stress
• Other non-contact sensors

On Friday, DCSA declined to provide additional information, including the scope of the "Modernizing Polygraph" project and the decision not to include AI-based voice screening and facial expression analysis.

Training Data and Methods

Over the summer, Farakh Zaman, a developmental leader at the Air Force Office of Scientific Research, elaborated on the appeal of AI-aided deception analysis:

"By processing subtle physiological, vocal and linguistic cues, these developmental algorithms may help establish a more objective baseline for investigators and pinpoint specific areas requiring further human-led clarification."

Wednesday's DCSA statement said the Air Force is not involved in the "active" research and development phase.

One undated declassified DCSA slide describes the Air Force amassing "deception-related" text to create a "deep learning model"—an algorithm that analyzes large datasets to find patterns. The Air Force trained an early model on text from:

• Twitter
• Reddit
• Other vocabulary datasets such as WordNet

"This baseline training helps the algorithms understand natural, modern linguistic patterns, slang and everyday vocabulary," Zaman explained in July. He acknowledged the need for "further rigorous testing, validation and a clear transition path" before fielding a prototype.

Avatar Interview System

A separate DCSA slide dated May 2025 contains a flowchart depicting an avatar—a computer-generated image of an interviewer—in front of a security clearance applicant. Next to the applicant, a camera and microphone feed audio, video, and speech-to-text data into various large language models. The recording devices and algorithms interact to gauge the applicant's emotional state, while a human investigator observes and inputs questions for the avatar to ask.

DCSA's Wednesday statement said the university that drafted the chart is not involved in the current research and development stage.

According to another undated DCSA slide, DoD has set a goal for natural language processing technology to interpret an interviewee's feelings with 75% accuracy.

In July, the Pentagon official said by email that a human investigator remains essential for "the contextual reasoning and emotional intelligence required in these sensitive interviews."

Expert Concerns

National Security Attorney Mark Zaid

Mark Zaid, a national security attorney who reviewed the slides, warned that the danger of derailing a military officer's career is too great for AI to misinform an investigator.

The polygraph device "is just registering the physiology of the person," such as spikes in breathing rate, perspiration, or blood pressure, he said. "Then a human examiner analyzes those reactions to render an opinion."

With AI, however, "AI would be taking the polygraph readings and rendering an opinion" to the human examiner, said Zaid, who has represented individuals on all sides of the polygraph table.

"Now, we're starting to get into the movie Minority Report," he said, referencing the 2002 Steven Spielberg film depicting a government reliant on psychic children to predict crimes and detain people accused of future offenses.

Zaid, whose own clearance was temporarily revoked after he represented a whistleblower pivotal in President Trump's first impeachment, said: "That concerns me more than the polygraph," if science does not back AI's reliability.

Scientific Evidence

So far, scientific studies do not support AI-based deception detection. Decades of research have discredited theories—popularized by shows such as Lie to Me—that anyone, human or artificial, can recognize deception based on speech patterns, vocal stress, or facial movements.

Bias Concerns

Another concern scientists and auditors have raised is bias. The Justice Department has warned that using emotion-recognition algorithms to measure worker competency can discriminate against high performers with intellectual or developmental disabilities, whom algorithms do not always understand.

Additionally, "automation bias"—the tendency to take AI's advice without question even when conflicting information exists—may further erode the polygraph system's reliability, technologists warned.

Legal Precedent

"There is simply no consensus that polygraph evidence is reliable," the Supreme Court declared in a 1998 decision banning polygraph results and polygraphers' opinions from military courts.

Conclusion

Zaid said he appreciates the attempt to retool polygraph testing but wants proof that the planned approach is in fact an improvement.

"Anytime we might be able to scientifically advance the ability to determine truth, with accuracy, would generally be a good thing," he said. But "the first thing that jumps out at me with AI is how often it is wrong… I've caught AI lying to me."