Pentagon denies using AI in polygraph loyalty analysis, says no generative AI or facial coding used

The Pentagon has clarified that certain artificial intelligence tools previously under development for analyzing trustworthiness are not part of the current overhaul of its polygraph interview process, following concerns raised by national security experts.

Background

As of July 2026, the Department of Defense had been preparing AI tools designed to analyze text, voices, and faces to estimate trustworthiness, according to declassified presentation slides and prior DoD statements. These efforts were part of a project referred to interchangeably as "Polygraph+" and "Credibility Assessment Modernization."

Current Position

On Wednesday, the Pentagon's Defense Counterintelligence and Security Agency (DCSA) issued a statement clarifying that the "Modernizing Polygraph" effort:

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

DCSA stated it will "continue to monitor emerging scientific capabilities to evaluate their potential utility down the road."

However, 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.

Contrast with Earlier Statements

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

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 were "primarily trained on data from laboratory-controlled studies using volunteer participants."

Project Scope and Funding

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

Training Data and Methods

DCSA slides 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.

One undated declassified slide describes the Air Force amassing "deception-related" text to create a "deep learning model." An early model was trained on text from:

• Twitter
• Reddit
• Other vocabulary data sets such as WordNet

Farakh Zaman, a developmental leader at the Air Force Office of Scientific Research, explained in July:

"This baseline training helps the algorithms understand natural, modern linguistic patterns, slang and everyday vocabulary."

He acknowledged the need for "further rigorous testing, validation and a clear transition path" before fielding a prototype.

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

Proposed Interview System

A DCSA slide dated May 2025 depicts a flowchart showing:

• An avatar (computer-generated interviewer) facing a security clearance applicant
• A camera and microphone feeding audio, video, and speech-to-text data into various large language models
• Recording devices and algorithms interacting to gauge the applicant's emotional state
• A human investigator observing and inputting 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.

Expert Concerns

Scientific Validity

National security attorney Mark Zaid, who reviewed the slides, warned that the stakes are too high for AI to misinform investigators:

"The polygraph device is just registering the physiology of the person... Then a human examiner analyzes those reactions to render an opinion. Here, however, AI would be taking the polygraph readings and rendering an opinion to the human examiner."

Zaid compared the approach to the film Minority Report, saying:

"That concerns me more than the polygraph," if science does not back AI's reliability.

Decades of research have discredited theories that anyone—human or artificial—can recognize deception based on speech patterns, vocal stress, or facial movements.

Bias and Automation Bias

Additional concerns raised by scientists and auditors include:

• Discrimination: The Justice Department has warned that emotion-recognition algorithms can discriminate against high performers with intellectual or developmental disabilities
• Automation bias: The tendency to accept AI advice without question, even when conflicting information exists, may further erode reliability

Legal Precedent

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

Zaid's Personal Experience

Zaid, whose own security clearance was temporarily revoked after representing a whistleblower pivotal in President Trump's first impeachment, said:

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

Mark Zaid, pictured in 2017