The potential clients for technology such as electronic jammers and interceptor drones tested at the Brétigny-sur-Orge site are no longer just governments, police and armies.

Private firms and state entities across Europe that operate critical infrastructure want to strengthen their defenses, with some utilities asking for authorization to shoot drones down themselves. From Paris to Berlin, governments are taking note.

They are worried that hybrid warfare or a drone attack could knock out power to millions of people — as has happened frequently in Ukraine after Russian strikes.

“It is a theoretical, and increasingly a practical, possibility that there could be heating outages this winter due to hybrid attacks,” Kristian Ruby, secretary general of European energy body Eurelectric, told Reuters.

Some European governments are responding by starting to give firms powers traditionally confined to the military. Interviews by Reuters with over 30 company executives, officials and drone experts meanwhile show the technology to detect and bring down drones is maturing faster than the rules on who may use it.

Approaches vary widely across Europe and legal uncertainties remain, including over who has the right to order a response to a drone threat. Bringing down a drone could cause injury or death to people on the ground or damage property, raising questions for companies about liability and accountability.

“Defenseless”

France is among countries that have led the way. Its parliament passed a law in August allowing approved operators of critical infrastructure and contractors to detect, jam and shoot down drones near designated facilities.

The law passed by the French parliament in August responded to a wave of drone incursions over critical and military sites, including flights over the Ile Longue nuclear submarine base near Brest in northwestern France in December 2025.

France wants a newly expanded list of sensitive sites deemed of vital importance equipped by 2027, a senior French official involved in the plans told Reuters, asking not to be named because details of the plan are not yet public.

“We were defenseless in the face of the (drone) threat,” the official said, adding that concerns had been growing over the last decade as unmanned aerial vehicles became more accessible and more common.

The French government wants to give operators of critical infrastructure such as state-owned utility company EDF and state-owned railway company SNCF a catalogue of approved systems ranging from electronic jammers to interceptor drones, depending on different environments and situations.

Confirming important elements of the plan outlined by the French official, Systematic Paris-Region, which runs the government testing program at Brétigny-sur-Orge, said evaluations launched this summer had so far catalogued three systems integrators — which combine different computer systems into a unified whole — and 17 counter-drone systems.

It said assessments were continuing and work was underway to accelerate procurement from among those solutions.

France’s defense ministry said the law passed in August would help strengthen capabilities to counter the threat, though it said it could speak only for the critical infrastructure under its authority. EDF and SNCF declined comment.

Who can bring down a drone?

The discovery of an explosives-laden drone near a Ukrainian cargo plane at Germany’s Leipzig/Halle airport in August underlined how far the risks have spread.

Germany blamed Russia for the incident. Moscow denied it was behind it, and has repeatedly denied carrying out hybrid attacks in Europe, many of which have been in countries that support Kyiv in Russia’s war in Ukraine.

But concerns are growing that national authorities cannot respond fast enough to protect widely dispersed infrastructure against such attacks, and that the damage caused by relatively cheap drones could bring hefty repair costs.

In Germany, where counter-drone operations remain a policing function, the interior ministry told Reuters it was reviewing how critical infrastructure operators and third parties could be legally involved in active counter-drone measures.

In September, Austria announced plans to give similar powers to critical infrastructure operators, including large energy firms. Defense Minister Klaudia Tanner told a press conference that the state could not protect every site on its own.

Poland has gone further. Operators told Reuters they have deployed counter-drone systems and trained with military and police units. Drone-restricted zones have been established around critical energy infrastructure, and designated security units working for operators have been authorized to destroy, disable or take control of drones entering them.

But two sources at energy transmission companies on NATO’s eastern flank told Reuters that effective intervention still requires more assurances, and perhaps a judicial precedent, on who bears responsibility for collateral damage.

Without that certainty, they said, operators of critical infrastructure could remain reluctant to neutralize drones because bringing one down could itself cause injury or costly damage.

Built for recovery

Having powers to bring down drones would be a change of approach for companies across Europe.

Executives, officials and energy analysts told Reuters that utilities have spent years hardening networks and preparing to restore disrupted services, rather than preventing attacks.

Leonhard Birnbaum, chief executive of German energy firm E.ON, said in September that operators must assume some attacks will succeed, rather than plan on stopping all of them. The company operates about 2,720 substations across Germany, he said.

Lithuania is building a fortified substation near its border with Russia as part of plans by the Baltic states and Poland to improve their defenses that include constructing bunkers, installing anti-drone nets and building up stocks of hard-to-replace equipment for critical parts of the energy grid.

Poland’s state-owned transmission operator PSE patrols parts of its about 16,500-km (10,250-mile) high-voltage network by helicopter and is preparing more powerful aircraft to help protect an undersea power link to Sweden.

Yet relatively inexpensive drones are challenging such defenses. Essential equipment such as large transformers can take years to replace, while even a successful repair can leave infrastructure out of service for months.

Not all utility operators are seeking a bigger role in defending themselves.

Asta Sihvonen-Punkka, chief executive of Finland’s national power grid operator Fingrid, told Reuters that responsibility for countering threats around critical infrastructure should remain with Finnish authorities and armed forces.

Any move to give such powers to utility operators would be “a very, very big change” requiring careful legal consideration, she said.

Jari Stenius, head of security of Finnish utility Fortum, said he was content for energy companies to focus on producing energy while police and military authorities handled security threats.

Technology ready, rules are not

Security officials and counter-drone technology firms say systems are becoming increasingly automated because operators cannot station specialist teams at every site.

Modern counter-drone systems combine sensors, command-and-control software and countermeasures into a single chain designed to identify, classify and respond to threats within seconds.

Several companies told Reuters that uncertainty over who can authorize a response has slowed investment in the latter stages of that process. France is therefore trying to automate the process as much as possible, said the senior French official involved in France’s plans.

The official said security workers at utilities should eventually be able to detect a threat and launch an automated response at the press of a button, because having operators trained to use military-grade solutions at every location was unrealistic.

German defense company Rheinmetall told Reuters the biggest hurdle for civilian operators was often the delay between detecting a drone and deciding whether to deploy a countermeasure, as legal responsibilities remain fragmented.

Rheinmetall said interest from private infrastructure operators in counter-drone systems was rising sharply.

The company said it was conducting workshops, site surveys and infrastructure assessments for operators and was preparing for what it described as a “very fast-growing” market segment.

(Reporting by Maria Rugamer and Leo Marchandon in Gdansk, Florence Loeve in Paris and Marek Strzelecki in Warsaw.)

Head of Army Cyber Command Lt. Gen. Christopher Eubank on Monday at the Association of the United States Army conference in Washington gave an approximate timeline for when the service would be delivering the new unit devoted to the exact same mission as the 11th Cyber Battalion, dubbed the Leviathans.

“It should start standing up in earnest within the next couple months, and it will be an exact replica of the 11th Cyber Warfare Battalion,” Eubank said.

The new battalion will achieve initial operating capability within the next year and from there work toward full operational capability, according to the Cyber Command leader.

The battalion is currently working on behalf of Joint Task Force–Southern Border, providing radio frequency-enabled cyber effects as directed by the division commander, Eubank said.

The Leviathans was activated on Oct. 16, 2022, as the first scalable expeditionary capability to handle the Army’s current and future cyberspace objectives.

It was built with the specific intention of rolling out non-lethal and non-kinetic effects that include cyberspace and electronic warfare assets.

Its previous iteration, the 915th Cyber Warfare Battalion, was activated on June 6, 2018, before being renamed during the 11th Cyber Battalion’s activation.

At the time of the battalion’s creation, the Army said it planned to have 12 expeditionary teams it could deploy under the 11th Cyber Battalion by September 2027.

The service has labeled the battalion the first of its kind.

“Just as Gen. Patton led the modernization of tank warfare into WWII, the leaders and soldiers of the 11th Cyber Battalion spearhead concept development for the delivery of CEMA [cyber electromagnetic activity] combat power to the tactical edge of the contemporary battlefield,” said Lt. Col. Lou Etienne, who led the 11th Cyber Battalion until June 25.

Developers from 11th Cyber Command worked alongside the Cyber Solutions Development unit within the 780th Military Intelligence Brigade to develop new drone technology between July 2025 and July 2026.

The partnership project was performed to support the Army’s “Transforming in Contact” initiative, which delivers new unmanned technology to soldiers to track their feedback in real time as they utilize it in the field.

The 11th Cyber Battalion also participated in a demonstration at Aberdeen Proving Ground, Maryland, from August. 5-8, simultaneously deploying three new Army technologies and proving their efficacy in diffusing electromagnetic threats from enemies.

Defense Secretary Pete Hegseth addressed AUSA attendees at the Project Agincourt panel on Wednesday, sharing that the Pentagon is requesting roughly $50 billion in reconciliation funding to support the push to scale autonomous and robotic systems.

This request is part of an effort to make the technology available in far greater numbers.

The figure is separate from the Pentagon’s fiscal 2027 budget request, which includes $53.6 billion for the Drone Dominance initiative. That request covers $16.9 billion for unmanned-system procurement, $14.4 billion for counter-unmanned systems, $13.5 billion for contested logistics, $4.5 billion for collaborative autonomy and $4.3 billion for institutional systems and personnel. The budget request does not establish that the reconciliation funding Hegseth cited is the same money or that it is dedicated entirely to drone purchases

Hegseth characterized the funding as a one-time investment rather than a recurring annual allocation, intended to accelerate the development and production of autonomous and robotic systems.

“Ultimately we will be dealing with hundreds of thousands of one-way attack systems this year and into the millions in years to come at a pace that this department’s never ever seen before.”

The Defense Secretary said the Pentagon must make it easier for new companies to compete in a defense market long dominated by established contractors and costly platforms. Traditional defense firms remain important, he said, but the department also needs lower-cost systems that can be manufactured at scale.

The Army’s Drone Dominance program offers a concrete measure of that ambition. In a December 2025 announcement, the Army said a $1 billion initiative would fund approximately 340,000 small unmanned aircraft systems for combat units over two years. The service said it expected to deliver tens of thousands of drones in 2026 and hundreds of thousands by 2027.

Acting Army Secretary Adam Telle said the service is broadening access to its programs for companies outside the traditional defense industrial base. The Army, he said, must make room for new entrants while continuing to work with established suppliers.

Telle also described plans to coordinate the Army’s autonomous-systems efforts while distributing capabilities to operational units. The service is developing an approach that incorporates drones, autonomous ground systems, autonomous fires and autonomous lift into future combat operations.

“We have an MOS in the Army, 390A, where warrant officers could keep the speciality and the skills over the duration to teach our soldiers how to do robotics,” Telle added.

Expanding access to Army test ranges is another part of that effort, Telle said. Greater access gives developers more opportunities to test systems and receive feedback from military users. He also described efforts to share lessons across the force rather than have individual units and organizations solve the same problems independently.

Hegseth said the department must remove bureaucratic barriers that slow the adoption of new technology, telling the panel, “I want you to tell me which rules can be signed away.” He urged officials to identify rules and directives that could be changed without involving new legislation to accelerate progress.

The service will pay a one-year total of $93.6 million, spread out across nine companies, to create applications for its Next Generation Command and Control, or NGC2, software.

“While the Army is providing capability we can deploy today, we’re also providing a blueprint and reference point for how industry can integrate in the future with interoperability and speed,” said Brig. Gen. Shane Taylor, portfolio management executive for the Command and Control Information Network.

The NGC2 will provide a shared network for technologies, weapons, sensors, devices and applications to operate within that will increase decision-making speed on the battlefield.

The chosen companies include General Dynamics Mission Systems, Air Space Intelligence Federal, Immersive Wisdom, LMI, Mente Systems, Stilman Advanced Strategies, Onebrief, Rune Technologies and AIR.

The applications will be hosted on the common data baseline, or software foundation, that Anduril Industries, Palantir, Raft and Striveworks have built in collaboration with the Army.

The NGC2 system, in conjunction with the applications the companies build for it, will allow information sharing that streamlines mission completion across the Army’s six warfighting functions of command and control, fires, intelligence, movement and maneuver, sustainment and protection, the service said.

The Army is currently in the process of delivering the NGC2 software to the I Corps, which commands service members in the Indo-Pacific.

The move will further test NGC2 software’s ability to host technologies and devices separated by thousands of miles, some in communication-contested environments.

The Army has spent the last year moving from the experimental phase to the fielding phase of NGC2, going from prototype development to placing software in the hands of soldiers at the 4th and 25th Infantry Divisions.

In July 2026, the Army hosted a 10-day event at Fort Irwin, California, entitled Project Convergence-Capstone 6, that asked 10,000 participants from U.S. and allied forces as well as industry partners using 90 technologies to test software for the NGC2.

The service conducted a joint force exercise shortly after in May at Fort Irwin that marked the end of the NGC2 prototype phase and came away stating that the new platform was ready to be scaled up.

The Army awarded Anduril Industries a $162.8 million contract on Sept. 30 worth up to 1.8 billion to continue refining the NGC2’s common data baseline, or software foundation.

Now, Pentagon officials say that some of that AI is not part of the current revamp of the polygraph interview process, after Defense News asked for a response to concerns about the earlier statements and the slides.

On Wednesday, the Pentagon’s Defense Counterintelligence and Security Agency, or DCSA, which vets security clearance holders and applicants, said in an emailed statement that “active research and development” for the “Modernizing Polygraph” effort “does not incorporate generative super intelligence, also known as artificial intelligence,” that creates new content — such as emails, graphics or voiceovers.

Nor does the effort include “large language models,” meaning chatbots like ChatGPT and other programs trained on giant data sets to engage in human conversation, DCSA added.

And the research does not involve “facial action coding [of emotions], micro-expression analysis or vocal analyses,” the statement said. “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.

Wednesday’s statement came after Defense News sought a response to criticisms from national security lawyers, scientists and AI auditors about AI’s ability to analyze deception — meaning not just to lie, but to do so knowingly and willfully.

They warned that AI-aided scanning of speech patterns, vocal stress or facial expressions to read minds lacks scientific backing, will compound errors and may violate civil liberties in a domain where few such rights exist.

Defense News chronicled their concerns and the Pentagon’s prior research and development of AI-aided lie-detection and emotion-estimation tools.

The recent DCSA statement, including the project name “Modernizing Polygraph,” contrasts with several earlier DoD statements, 2027 budget documents and DCSA slides that Defense News obtained through an open records request.

In July, a Pentagon official, who spoke on condition of anonymity to discuss the technology, explained the advantages of AI-driven sentiment and deception analysis.

“AI processes data in real time and enables standardized, objective data analysis,” the official told Defense News, via email. “AI-based analysis is a force multiplier for our human investigators, not a replacement.”

The official added, “The models are primarily trained on data from laboratory-controlled studies using volunteer participants.”

The DCSA slides, which include 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,” which the official referred to interchangeably as “Polygraph+” and Credibility Assessment Modernization.

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 spot blood flow changes that can reflect stress and 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.

Trained on Twitter and Reddit

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,” he said by email in July.

Wednesday’s DCSA statement said that 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 hordes of data to find patterns. The Air Force trained an early model on text from websites including Twitter, Reddit and other vocabulary data sets such as WordNet, according to the slide.

“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.

A separate DCSA slide, dated May 2025, contains a flowchart that depicts 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 from the interview into various large language models. The recording devices and algorithms interact to gauge the applicant’s emotional state, while a human investigator observes and keys in 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.”

‘I’ve caught AI lying to me’

Ahead of Wednesday’s statement, 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.”

Here, 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, from workers disputing revoked clearances to former government officials and polygraphers.

“Now, we’re starting to get into the movie Minority Report,” he said, referencing a 2002 Steven Spielberg film that depicts a government reliant on psychic children for tips to predict crimes and detain people accused of future crimes.

Zaid, whose own clearance was revoked temporarily 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.

So far, scientific studies do not.

Rather, 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.

Another concern scientists and auditors have 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, who algorithms do not always understand.

Also, “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.

“[T]here 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.

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.”

Greg Rinckey, a former Army Judge Advocate General Corps officer who now provides security clearance representation, said that today’s “technology is so outdated, where they’re using tubes around people’s chests and blood pressure and sweat, so, any way that you can get a more reliable test for security clearances, especially counterintelligence investigations [into leaks], is a step up.”

That said, he questioned, “What is the research saying on how accurate it is?”

At present, research suggests the odds of AI-aided speech, voice or face analysis pegging a lie are about 50:50.

AI is ‘assuming dishonesty’

“There are no – reliable – diagnostic cues for deception detection, and I have yet to see any evidence, whether it is with humans or with artificial intelligence, that goes against that trend,” said David Markowitz, a Michigan State University professor who studies AI’s ability to understand human communication.

For instance, “the leakage idea — that these microfacial expressions will leak out from our awareness, signaling veracity or deceit — there’s very little evidence to support that, unfortunately,” he said.

Markowitz expressed doubts about AI-based deception analyses that claim an accuracy rate of 75% because “that is not how accurately humans do it, and often these systems are a big reflection of how humans do it.”

For example, he pointed to a recent experiment in which he prompted a Google Gemini AI model to examine audio and video of mock interrogations of accused cheaters in a trivia game, all of whom claimed innocence and some of whom were lying. Prior studies had asked humans to observe the same interrogations. DoD has entered into agreements to run a secure version of Gemini on unclassified and classified networks.

Markowitz said that his study, “The [in]efficacy of AI personas in deception detection experiments,” found Gemini was just as inaccurate as humans, both faring no better than chance, when prompted to distinguish fraudsters from truth-tellers.

What’s more, compared to humans, AI was more likely to wrongly accuse a trivia player of lying.

Markowitz gave one possible reason for AI’s systemic distrust of trivia players: The AI “is assuming that someone must have done something wrong because why else would they be questioned,” he said. “Anytime that someone is in an interview setting where they might be accused of something, a large language model is assuming dishonesty.”

Supreme Court: The jury is the lie detector

The consequences of a machine, whether a heart monitor or a voice screener, potentially bungling a credibility assessment can last a year or longer, said Zaid, the national security lawyer.

When officers who have successfully passed exams without issue suddenly fail their latest polygraph test, DoD typically puts them “in a corner with a dunce cap on for a year before officials will re-administer another exam,” he said. “You can’t get promoted. You can’t get certain assignments. And you can’t go overseas, which, to case officers, is death careerwise.”

Further, civil courts routinely dismiss complaints when employees claim that inaccurate results cost them their clearances. Nearly 40 years ago, the Supreme Court ruled that the courts have no power over such national security matters.

Concerns about power imbalances and due process prompted the European Union to label AI-aided polygraph tools as “high risk” and place strict limits on law enforcement’s use of them.

Stateside, the Supreme Court, in forbidding polygraph evidence in military courts, noted that allowing it would meddle with the human’s role, or, as it were, the jury’s role, in determining credibility.

“A fundamental premise of our criminal trial system is that the jury is the lie detector,” Justice Clarence Thomas wrote in the Court’s 1998 United States v. Scheffer decision.

Pointing to the court’s decision, Zaid said that he wants assurances of due process when staff undergo administrative proceedings to contest a revoked clearance resulting from misaligned algorithms. Comparing the situation to successful cases disputing speeding tickets caused by a miscalibrated street camera, Zaid questioned, “Would the person be provided the evidence? An understanding of the AI’s training?”

New biases

Pentagon officials have long stressed the desire for AI-based lie detection to help curb gender and cultural biases that human investigators may carry into an interview room.

At the same time, employers’ use of similar technology for pre-employment screening has faced allegations of other biases and system errors.

“Facial expression analysis is heaviest in terms of biases against individuals with disabilities,” said Jo Ann Oravec, a University of Wisconsin professor whose research focuses on AI, public policy and disability issues.

For instance, AI-powered video screening of job applicants tends to misinterpret the involuntary twitches or vocal outbursts of skilled people with certain disabilities — and thus pass over qualified applicants, said Oravec, author of the 2024 article “From Polygraphs to Truth Machines: Artificial Intelligence in Lie Detection.”

In response to concerns about such biases in July, the Pentagon official said via email that the credibility assessment modernization plan adheres to DoD’s AI Ethical Principles of “responsible, equitable and traceable” use of algorithms.

The Air Force’s Zaman, in his July email, said that his team’s work aligns with DoD’s Responsible Artificial Intelligence Guidelines.

Researchers “are actively engineering and testing these models” to account for “diverse, representative training data,” he stated. The testing covers “broad variations in speech, language and expression.”

Ryan Carrier, founder of ForHumanity, a global nonprofit that helps organizations audit AI systems, said that the Pentagon’s principles and guidelines are a skeleton of those that the European Union follows and that ForHumanity uses for inspections.

For example, the EU’s AI Act demands users of high-risk systems receive training to overcome a different kind of bias called “automation bias” — or the tendency to execute AI’s advice without questioning its accuracy. This bad habit can lead to incorrect data overriding correct decisions, Carrier said.

Also, multimodal systems, meaning AI systems that simultaneously parse audio (the subject’s voice), video (their facial movements) and text (interview transcriptions), tend to misalign data and results, yielding an irrelevant analysis, Carrier said.

The system “could be making determinations that a person is lying or trying to deceive me that are outside of the tool’s specifications,” he explained.

In response to concerns about multimodal systems in July, the Pentagon official had said, “Any automated output is treated strictly as auxiliary data for a highly trained, certified polygraph examiner.”

Similarly, Zaman, at the time, said that investigators who use AI-aided deception-detection systems are “trained to contextualize the tool’s findings, preventing automated outputs from dictating an outcome.”

To safeguard against mismatched data or results, the Air Force keeps a documented history of each piece of data’s origin and modifications and brings in independent white-hat hackers to try producing false correlations, he added.

“We recognize that multimodal AI systems present unique technical challenges, and mitigating these risks by hardening our systems is central to our engineering and testing protocols,” Zaman said.

Reason for suspicion?

In August, when the Pentagon administered a raft of polygraph exams to probe leaks alleging the Iran conflict is depleting U.S. arsenals, some observers suspected the endgame was not to find the leakers, but to crack down on disloyalty.

Critics of the planned system said that AI-based deception analysis may be best suited to achieve the same purpose.

“If it’s purely an intimidation technique, it may not be utterly useless from the point of view of large bureaucracies,” said Jay Stanley, a senior policy analyst with the American Civil Liberties Union’s Speech, Privacy and Technology Project.

The expectations for AI-powered trustworthiness testing resemble the hope that U.S. national security officials held out for the polygraph about 100 years ago, other lawyers said.

Polygraph devices “were seen as this magic-miracle-machine when they first came out,” and, over time, the government learned that when seemingly making deductions, the tools were “actually just flagging correlating factors, like heart rate and physiological factors that can indicate lying, but aren’t actually reflective of it,” said Jake Laperruque, deputy director of the Center for Democracy and Technology’s Security and Surveillance Project.

He added, “Thinking that AI sentiment analysis tools are actually making an assessment or engaged in reasoning — that’s not what those machines are actually able to do.”

Speaking Oct. 1 at the Bergen military training area during the army’s “Experimentalserie Land” trials, Pistorius said ground forces will begin forming the regiment in 2027. Officials credited direct operational exchanges with Ukrainian frontline units for shaping the live-fire trials, which tested both aerial drones and unmanned ground systems. Pistorius emphasized that ground robotics will serve as an equal operational pillar alongside aerial systems, with the regiment consolidating reconnaissance, deep-strike capabilities, unmanned ground vehicles, electronic warfare, and counter-drone defenses under a unified command.

The regiment will serve as a divisional asset assigned to the 10th Panzer Division − the formation earmarked as Germany’s high-readiness “Division 2025” pledge to NATO, which also oversees Panzerbrigade 45 in Lithuania.

German Army Chief Lt. Gen. Christian Freuding said the Bundeswehr is pursuing a dual-track strategy that envisions setting aside dedication formations while also institutionalizing drone usage throughout the entire force.. At the turn of the year, the service will simultaneously begin fielding tactical unmanned systems directly across three core combat branches: light infantry, reconnaissance, and armor units.

The army plans to recruit between 300 and 350 personnel for the regiment’s initial footprint in 2027 before expanding the unit to an end strength of up to 600 troops, according to Freuding. In remarks to German wire service dpa, Freuding said planners target full operational capability by 2029.

The envisioned structure comprises five or six companies, including units for close-range combat, deep strikes at ranges of up to 500 kilometers depending on commercial equipment availability, electronic warfare, counter-UAS protection, and an internal trials company intended to fast-track operational testing. The Bundeswehr also plans to equip the formation with loitering munitions, which are currently undergoing evaluation and should reach troops in 2027, according to dpa information.

Unmanned systems have fundamentally reshaped modern combat, accounting for the vast majority of casualties on the battlefield in Ukraine. Western militaries have often lagged in retrofitting their forces, slowed by peacetime procurement practices and protracted political debates over weaponized drones and autonomy.

Key procurement questions surrounding the new unmanned-tech formation remain. The defense ministry has yet to finalize a permanent basing location or award contracts for the specific long-range platforms the regiment requires.

The project seeks to develop open, modular architecture for combat aircraft and sovereign AI technology that gives the government greater control over the tech, rather than having to rely on companies for critical systems, said Col. Cédric, head of the Air Force capability-development office, in a press briefing here on Thursday.

As AI becomes more powerful, the likes of Airbus and BAE Systems in Europe and General Atomics and Anduril in the United States are leveraging the technology to develop collaborative combat aircraft, also referred to as “loyal wingmen.” While many of those projects are company-driven, the French project would keep some of the critical technology under state control.

“What Hypairion is really about is equipping the Air Force and our platforms with architectures that are open, modular and owned by the government,” Col. Cédric said. He was identified only by his first name, in line with the Armed Forces Ministry’s policy for military staff. “Our goal is to move away from proprietary systems that lock down sensors and weaponry.”

The Ministerial Agency for Defence AI will be in charge of providing sovereign AI technology for the project in several areas, particularly command and control as well as mission autonomy, said Col. Thibault, deputy director for AI applications at the agency, known as Amiad. He described mission autonomy as AI that assists pilots with mission planning, in-flight decisions and post-mission analysis.

Amiad already uses AI to coordinate different platforms in the French Army’s Pendragon robotics project, and proposes to create an industrial hub for Hypairion where both traditional aerospace firms and smaller startups can interact with the project, an approach tested in Pendragon.

“The point of the Hypairion project is for the French state to retain control of the overall architecture and to be able to integrate the most capable startups or industrial companies,” Col. Thibault said. He said companies might join and leave the project, while Hypairion retains its capabilities in-house.

The Air Force is modifying two Mirage 2000s fighters at Mont-de-Marsan airbase to become AI test beds, adding a new computer that will allow to test “all kinds of use cases” starting by the end of the year, according to Col. Cédric. Mont-de-Marsan in southwest France is home to the Escadron de chasse et d’expérimentation, a dedicated fighter squadron for testing and evaluation.

Development will be incremental, and the first stage of Hypairion won’t necessarily be about having an operational capability, but will look at areas such as energy consumption and system interfacing, Col. Cédric said. First use cases being considered include using AI to optimize flight paths or tactical decisions such as when to deploy air-to-air or air-to-ground weapons.

The unmanned aircraft set to fly in 2028 will be a test platform and may not be the future wing-man drone planned for France’s Rafale jet as part of the fighter’s future F5 standard, though the experimentation will help pick a suitable platform later in the decade, Col. Cédric said.

The DGA has set a target data of 2033 for the first version of the Rafale F5, but will work on collaborative platforms within the framework of the current F4 standard before then, said General Engineer of Armaments Sébastien, in charge of the Rafale program at the directorate.

The goal of integrating AI on the Rafale is to help crews manage the “absolutely phenomenal cognitive load” to maintain future operational superiority, he said. The armaments engineer cited the Talios laser-designation pod from Thales, set to be delivered to the armed forces in 2027, as a concrete application of AI, able to scan an area and provide a fighter crew a list of potential targets.

Col. Cédric described a hypothetical future in which a fighter crew would carry out a mission in the morning, in between missions identify an adaptation against an enemy countermeasure, and take off again in the afternoon with updated software. That will require system architectures to evolve, he said.

The French Air Force officer likened future manned and unmanned platforms to “flying computers” on which the keyboard, screen or software could be changed without compromising security. The goal of the architecture being developed within Hypairion is to separate the flight-safety aspect from the ability to carry out missions, he said.

The Air Force doesn’t envisage a future with only unmanned aircraft, for both ethical and operational reasons, according to the colonel, who said the decision to deploy weapons remains a human one, and AI can make mistakes. “Humans can be augmented or supplemented by AI, but the assessment of the operational situation, tactical judgment, and so on, these are still entirely the domain of humans.”

As the Air Force prepares the Rafale F5 standard, “a true mid-life overhaul of the aircraft,” focus areas include adequacy of hardware resources, as the system needs to be able to handle complex calculations, Sébastien at the DGA said. Another focus is system openness, meaning “the armed forces must be able to tailor artificial intelligence according to their own needs and so dispose of an open system.”

The Rafale’s F4 standard already includes an initial level of openness through a tablet connected to the mission system, and the F5 standard will seek to broaden the scope of data available to the tablet, give it access to the mission system’s computing resources, and improve the ability to interface with onboard communication systems and data links, according to the DGA engineer.

“This in particular will make it possible, and without waiting for the deployment of the F5 standard, to control autonomous combat platforms from the Rafale,” Sébastien said.

While France relies on Dassault Aviation and other major industrial partners for the Rafale to develop AI in the fighter’s major systems, the armed forces want to ensure operational autonomy, meaning having “the ability to tailor the AI to the specific operational needs of the theater,” the engineer said.

Hypairion will work on a road map covering usage concepts, doctrine, scaling up the work force as well as the actual equipment, resulting in a first full-scale trial of collaborative combat drones flying alongside fighter jets in 2028, according to Col. Cédric.

“The challenge for the Air and Space Force is ultimately to determine what a fighter squadron will look like in 2035 or 2040, once all these systems are potentially integrated into the squadron,” Col. Cédric said. “Today, it’s actually quite rare to come across computer scientists in a fighter squadron. Perhaps tomorrow, that will be a daily occurrence.”

According to the competition’s leaderboard, Neros Technologies scored highest in the confined-environment testing category. Perennial Autonomy claimed the top spot in “deep-strike” testing, in which operators launched attacks beyond the range of enemy defenses.

The Pentagon uses the companies’ rankings to determine how many drones to buy, meaning higher-scoring companies receive larger orders. In total, the department ordered 14,000 drones from Neros — which also ranked third in deep-strike testing — and 8,000 from Perennial.

Neros, founded by competitive drone racers in 2023, identified its winning design as the Archer AI drone. The company said the 10-inch model was “built with Terminal Guidance functionality that allows the drone to fly itself to an operator-selected target, a critical feature in the comms-denied environments military operators face today.”

In a statement, Soren Monroe-Anderson, Neros’ chief executive and co-founder, called the Archer “highly modular and easily adapted to end-user needs.”

“Pairing real warfighter feedback with our deep engineering talent is the reason Neros is the only manufacturer to have placed in the top three across two fundamentally different mission sets — deep strike and close-quarters battle — with the same drone platform,” he said.

The Defense Department launched Drone Dominance in December 2025, with plans to invest $1 billion directly into the domestic drone market and build up the military’s supply of commercial drones.

Since then, the Pentagon has completed one-and-a-half phases, known as “Gauntlets,” along with three operational tests. Neros has ranked second, third and first in those competitions, earning the company roughly $100 million in orders.

Perennial said its design was “tested across rigorous find-fix-finish objectives at various long-range beyond visual line-of-sight (BVLOS) distances” and flown by operators during day and night missions.

“The systems proved their operational reliability under heavily contested conditions, to deliver precise, autonomous engagement when it counted,” the company said in a press release.

Earlier this year, the Defense Department awarded Perennial a $500 million contract through Joint Interagency Task Force 401, or JIATF-401, which is responsible for researching, testing and procuring counter-drone technology for the department.

The company was founded in 2023 by former Google CEO Eric Schmidt and developed the Merops interceptor for Ukrainian forces to counter Russia’s one-way attack drones.

Beginning in October, the Drone Dominance competition will enter Gauntlet 2.5, in which officials will test bomber platforms. The third phase is expected to begin next spring, with the competition scheduled to conclude next fall.

Then he played an AI-generated clip of what he meant: a robopunk mix of humanoid and crab-like ground robots skittering across a barren front line, doing work that is still too complicated for the robots Ukraine fields today.

The video laid out Fedorov’s vision for an Army of Robots, a force of two-legged upright and other ground robots that would take over many of the front line’s most dangerous but necessary jobs from soldiers.

“Our idea is to combine engineers, startups and capital into one ecosystem so we can launch humanoid-like robots on the battlefield,” Fedorov told Defense News in an exclusive interview after his presentation.

“We want to eventually substitute our infantry with these robots.”

To get there, Fedorov is calling on specialists in robotics, AI, sensors and autonomous systems to pool their talent in the kind of concentrated wartime push that built the first atomic bomb, and to focus it on the machines he says will one day replace Ukraine’s infantry.

The 35-year-old aims for the group to have its first confirmed battlefield result from a humanoid-like robot within six months and, within a year, an assault in which robots take on the most dangerous parts of a mission with no infantry alongside them.

To hit the six-month mark, there isn’t time to spend on research and development, he said, so he will begin by investing in companies with weapons that have already been tested and shown some market viability.

Fedorov wants Ukraine to seize the initiative in the war’s next technological cycle before Russia – whose far larger army and industrial base let it scale new weapons quickly – gets there first.

“Russia always thinks about new strategies, how to hack our frontlines,” he said. “We need to be the first side to hack our frontlines.”

Robots are only one of the defense tech areas that the former minister plans to develop within the new project.

“It could be robots, it could be autonomous systems,” he said. “We will have three key directions: sensors and robots, cheaper anti-jet-Shahed missiles, and AI strike drones and missiles.”

Under Fedorov’s plan, members of the Army of Robots will identify a battlefield problem, find a technical solution and a team, provide resources for development, test the product with troops in real conditions and scale what works.

It is the same model he used for the Army of Drones, which he started in 2022 as Ukraine’s digital transformation minister and which opened the market to Ukrainian drone makers, launched government procurement and operator training and helped form the first strike drone units.

“The Army of Drones … changed everything in our country and in the world and, I think, in the future of the world,” he said. “Now we are trying to build the same ecosystem for robots.”

Ukrainian troops already lean heavily on ground robots for assaults and basic tasks, carrying out more than 25,000 logistics and evacuation missions in August – 3.3 times more than in January, according to the Defense Ministry.

But most of the futuristic robot designs featured in Fedorov’s video, humanoids included, are not yet used in combat in Ukraine today, and the ground robots that are in use almost always have a human operator at the controls, usually over Starlink.

Fedorov said the ambitious effort will involve multiple companies working on different stages of the product lifecycle, with some making components, some assembling complete robots and others developing software to run them.

“We have some funds for these first companies,” Fedorov said, declining to name names. “It means that we have two or three companies that will be our partners in these directions.”

The Army of Robots will include a venture fund to invest in technology teams, its own defense-tech companies and a research and development arm, according to a press statement issued after the announcement.

The initiative comes weeks after Fedorov announced that he was starting a new defense company, with Palantir CEO Alex Karp as its first investor.

Fedorov told Defense News the company will start with a corporate fund and may later launch its own venture fund.

The biggest challenge facing the development and deployment of these next-gen robots is the necessary speed of innovation, Fedorov said.

Ukraine has to field these robots fast enough to save its infantry and stay ahead of Russia, which he said has learned to copy Ukraine’s battlefield innovations and scale them rapidly.

Ukraine, meanwhile, is heading into another winter of Russian strikes with air-defense interceptors in short supply.

The weekend Fedorov made his announcement, Russian drones hit kindergartens and university buildings in Kyiv, struck a large civilian hospital and destroyed a food market and multiple gas stations outside the city, killing at least three people and injuring dozens in a combined aerial attack that lasted days.

“We don’t have enough time,” he said. “We need to achieve these results as soon as possible.”

The 2-1 decision by the U.S. Court of Appeals in Washington came in a lawsuit by Anthropic challenging its March designation by the Pentagon as a national security supply chain risk, which the startup says has cost it billions of dollars in lost business and damaged its reputation ahead of a highly anticipated initial public offering.

The court’s majority said it was reasonable for the Pentagon to designate Anthropic after it refused to allow its products to be used for autonomous weapons or mass surveillance. It rejected Anthropic’s claim that the Pentagon had retaliated against it for its views on AI safety and ethics.

Siding with Hegseth’s argument that Anthropic’s safety restrictions could compromise military operations, the majority said, “The Secretary raises the deeply sobering prospect of overly constrained AI models shutting down unexpectedly and thus causing important military operations to fail,” in an opinion by Judge Gregory Katsas.

San Francisco-based Anthropic said in a statement Friday that it respectfully disagrees with the decision but remains confident in its position and is considering its options, including review of the three-judge panel’s decision by the full appeals court.

The White House and the Department of Defense did not immediately respond to emails seeking comment.

A federal judge in San Francisco last month struck down a parallel designation under a different law, finding the administration had unlawfully retaliated against Anthropic for its views on AI safety. U.S. District Judge Rita Lin in San Francisco also blocked Trump’s government-wide ban on Anthropic and Hegseth’s order barring military contractors from doing any business with the startup.

Hegseth designated Anthropic as a risk under two different laws after Anthropic refused to remove guardrails against its technology being used for autonomous weapons or mass domestic surveillance.

The unprecedented move came after months of negotiations ended in failure and public recriminations by Hegseth and Trump, who accused Anthropic of endangering American lives with ideological posturing.

Anthropic denied that characterization and said AI is not yet reliable enough to be safely used in autonomous weapons. The company also said it opposes domestic surveillance as a violation of fundamental rights.