Defense One - Science & Tech

‘Best drone’ innovation winner developing enemy drone recovery system with the Army Research Lab

Meghann Myers — Thu, 16 Apr 2026 18:01:24 -0400

NASHVILLE—A group of soldiers from the Pennsylvania Army National Guard is teaming up with the Army Research Laboratory to develop a prototype enemy drone recovery system that won the innovation award at the Army’s first Best Drone Warfighter competition in February.

The idea came together over a “couple beverages” after Chief Warrant Officer 3 Robert Reed and his 28th Infantry Division Innovation Team got the invitation to enter the competition late last year, he told an audience Thursday at the Army Aviation Warfighter Summit.

“So we wanted to come up with something that wasn't just the run-of-the-mill, Army-type system, something that industry would be excited about and potentially be able to take and make it scale from there,” he said.

They came up with Project RED—Recovery Exploitation Drone—an unmanned system that uses AI to find downed enemy drones, and an attached robot arm to pick up those drones and fly them back to the unit to download their data.

“We're currently working with Army Research Laboratory at this time to kind of refine our product, create more autonomy, more stability in the flight controls,” Reed said. It’s part of a one-year research-and-development agreement with ARL.

Other units are already starting to work on their pitches for next year, Reed said as part of a panel discussing lessons learned from the first Best Drone Warfighter competition.

He suggested creating sub-categories for the “best innovation” award with different budget thresholds, to give units of varying size and resources more room to develop their ideas.

He was joined by Sgt. Javon Purchner and Staff Sgt. Angel Caliz, who won the best drone operator and best team portions of the competition.

Purchner, a fire support specialist, brought several years as a first-person view drone hobbyist to the competition, he said. He suggested units give soldiers more designated time to train on drones.

“At installations, have actual courses for soldiers that want to compete,” he said. “They can have time to actually go out and practice their flying skills and have that time set aside for them, because flying FPV drones isn't just as easy as picking up the controller and flying. It's something that takes a lot of time and practice to become proficient at.”

Purchner’s leaders were so impressed with his skills that they plucked him from his unit in 1st Cavalry Division to serve at III Corps headquarters and develop a training center with multiple levels of courses to train new drone pilots at Fort Hood, Texas.

Caliz said he’d been practicing the hunter-killer drone mission with his fellow 2nd Cavalry Regiment soldiers, giving him an edge.

He suggested that next year’s competition include electronic warfare interference to make the scenario more realistic. He also had some suggestions for industry to make the drone mission more viable.

One would be some sort of transport system for drones, he said, because when you’re a team carrying five to 10 of them, they no longer just fit in a backpack. Another idea was a new kind of ground control station.

“I'd like to see more mobility. Something smaller, more compact,” he said. “Something that doesn’t tie you down to a certain case or certain bag, something that’s not too many wires.”

Assuming the second annual Best Drone Warfighter competition becomes a reality, leadership would like to expand the challenges with a nighttime portion, said Col. James Brant, the lessons learned and training manager for the Army Aviation Center of Excellence.

]]>

Russians will surrender to robots. Russian robots won’t.

Patrick Tucker — Wed, 15 Apr 2026 22:19:14 -0400

NATO is studying how to use ground and air robots to replace human soldiers in assaults, something Ukraine has been doing for more than a year. But that hasn’t stopped Russia’s continuous assault with its own, increasingly autonomous one-way attack drones.

On Tuesday, Ukrainian President Volodymyr Zelensky made a social-media splash with a video describing a historic first from last July: a skirmish in which Russian troops surrendered to Ukrainian robots.

“The future is already on the front line—and Ukraine is building it,” Zelenskyy said in the video, adding that Ukrainian robotics companies “have already carried out more than 22,000 missions on the front in just three months.”

Still, the Ukrainian president offered far fewer details than did Ukraine’s 3rd Assault Brigade in its own July 2025 post.

“Enemy fortifications were attacked” by first-person-view aerial drones and ground robots armed with explosives and made by Nazemnyi Robotychnyi Kompleks, the post said. “The next robot was already approaching the destroyed dugout when the enemy, in order to avoid being blown up, announced surrender. The occupiers who survived were taken to our lines by ‘birds’ [aerial drones] and, according to the regulations, taken prisoner.”

“The operation was carried out without infantry and without losses on our side,” it said. “The occupiers surrendered to the ground robots of the Third Assault!”

Ukraine’s ground-robot game advanced quickly in the following months, said Olena Kryzshanivska, a senior editor at the NATO Association of Canada who first relayed the news to English-language audiences.

“Already…[by the] beginning of this year, we saw several documented cases when UGVs [unmanned ground vehicles] were used for strike missions. They were either delivering grenades [or] they were sometimes … attacking trenches, attacking Russian troops,” Kryzshanivska said in February during a podcast with CNAS adjunct senior fellow Sam Bendett.

That sort of combined robotic fast maneuver is one of the ways Ukraine is forcing a reconsideration of decades of military doctrine, and NATO is taking notice. In February, its Allied Command Transformation announced the extension of a study on Force Lethality Enhancement to build out “a few practical force options and test them against realistic scenarios to see what works, and what it would take to use them on operations.”

Another alliance effort to integrate ground robots, part of the multidomain Task Force X, is being led by Brig. Gen. Chris Gent, NATO deputy chief of staff transformation and integration.

Venture capitalists are taking note as well. Eric Brock of Ondas Capital told Defense One in January that his firm is investing in “ground robots that are tailored towards defense and homeland security but also critical infrastructure protection in certain places.”

Challenges

The biggest constraint in using first-person-view drones is that an operator can generally fly just one at a time. But the drone can fly itself to waypoints, loiter in the air, and reconnect after brief communications interruptions.

Ground robots, by contrast, need constant attention because navigation remains a technical challenge, John Hardie, of the Foundation for the Defense of Democracies, told reporters in February. And UGV operators must also stay in frequent contact with the operators of the aerial drones above.

“My understanding is that they've experimented with autonomous navigation, but it’s especially difficult with [unmanned ground vehicles] for that to be reliable. So I don't think they're there yet,” Hardie said.

Ukraine has also been hunting for alternatives to GPS, which is jammable. Since 2023, it has been experimenting with visual- and terrain-matching systems and other AI-powered ideas for long-range navigation, Hardie said.

Russia, too, has carried out robotic operations in large volumes. But they’re limited to strikes with one-way attack drones like Shaheds and, occasionally evacuation of the wounded, not taking positions.

The Lancet drones produced by Russia’s ZALA company are guided on final approach to their targets by matching camera imagery to preloaded maps. It works well enough—because Russian forces place less of a premium on collateral damage or striking the right target.

For Ukrainians, the goal is greater autonomy, allowing one operator to preside over fleets of ground and air robots but with confidence that they will perform the mission assigned, hit the target that they’re supposed to hit and not simply whatever happens to be there when the drone finally arrives. It’s the same sort of complex multi-drone swarm capability that the Pentagon is seeking to develop.

Ukrainian Air Force Capt. Max Maslii, deputy chief of staff for the 96th Anti-Aircraft Missile Brigade, described that goal as a departure from the way Russia operates “autonomous” drones like the Lancet, as isolated flying bombs.

Under the “new paradigm,” Maslii told Defense One, the drones would be able to “find the … more efficient way to accomplish this mission, together with such machines.” At that point, he said, operators wouldn’t be stuck piloting one drone at a time. They would work more like technicians managing a larger, more complex system.

“Our job will be … to produce a lot of drones, to put them in the proper place, to take care [of] the systems that manage those drones, and just to, you know, turn them on.”

]]>

Put nuclear reactors in space within a few years, White House tells Pentagon

Thomas Novelly — Tue, 14 Apr 2026 18:12:22 -0400

COLORADO SPRINGS, Colo.— Launch nuclear reactors to orbit as soon as 2028 and to the Moon as soon as 2030—that’s the White House’s new order to the Pentagon and NASA.

The six-page policy memo released on Tuesday calls for a dual design competition between the agencies that is to produce a “nearterm demonstration and use of low- to mid-power space reactors in orbit and on the lunar surface.”

“The United States will lead the world in developing and deploying space nuclear power for exploration, commerce, and defense,” the policy reads. “Agencies will establish cost-effective partnerships with private-sector innovators to meet near-term objectives that include safely deploying nuclear reactors in orbit as early as 2028 and on the Moon as early as 2030. Achieving these near-term objectives will establish technological viability essential to unlocking space exploration, commerce, and defense applications.”

Michael Kratsios, the director of the White House’s science and technology policy office, unveiled the policy at the Space Symposium here. He tied it to President Donald Trump’s December executive order that aimed to “ensure space superiority” for the United States.

“Nuclear power in space will give us the sustained electricity, heating and propulsion essential to a permanent robotic and eventually human presence on the moon, on Mars, and beyond,” Kratsios said.

The defense applications for a nuclear reactor are wide-ranging, said Todd Harrison, a space policy and budget expert for the American Enterprise Institute. With a reliable energy source, the military could use it to power some of its most crucial future missions.

“You could run data centers in space, you could use it to power mission-critical systems that can never really go without power, like missile warning, strategic communications, Harrision said. “Directed energy, jamming, data centers, all of those things could use a lot of power.”

Within 90 days, the Pentagon must brief the White House’s science and technology policy office, management and budget office, and National Security Council on “relevant use-cases and payloads” for the systems and “best use of the 2031 mission,” according to the policy.

Those offices, along with the Defense Department, will decide on the final mission for that technology.

On Earth, the Defense Department has worked for decades to field nuclear microreactors to power its military bases. Last year, the Army announced last year that it aimed to break ground on a microreactor on a U.S. base by 2027. As well, the Pentagon’s Defense Innovation Unit declared eight companies eligible to build those microreactors.

Last week, the Air Force and Defense Innovation Unit selected Buckley Space Force Base, Colorado, and Malmstrom Air Force Base, Montana, as possible locations for two microreactors. There is also a standalone pilot program that will test the operational benefits of a reactor at Eielson Air Force Base, Alaska.

Top U.S. officials have dismissed the fears of groups such as the Bulletin of Atomic Scientists, who have pointed out that microreactors on U.S. bases “could become attractive targets for an adversary.”

Currently, there are no nuclear reactors in space and no operational microreactors on Earth within the United States. Harrison said the White House’s timeline for moon-based reactors is ambitious.

“The timeline and feasibility strikes me as rather aggressive,” Harrison said. “Demonstrating a microreactor on Earth would be challenging by 2028, doing it in space is even more challenging.”

]]>

CIA employees will get AI 'coworkers'—and eventually run teams of AI agents, deputy says

David DiMolfetta — Thu, 09 Apr 2026 14:45:00 -0400

The Central Intelligence Agency aims to integrate artificial intelligence-powered “coworkers” into analysts’ workflows in coming years, a top official said Thursday.

CIA Deputy Director Michael Ellis said these AI coworkers would be housed in agency analytics platforms to help humans with basic tasks.

“It won’t do the thinking for our analysts, but it will help draft key judgments, edit for clarity and compare drafts against tradecraft standards,” Ellis said in a speech at a Special Competitive Studies Project event on AI and the intelligence community. The AI tools would help triage and flag trends for human analysts to review.

And within a decade, Ellis said, the CIA will treat AI tools as an “autonomous mission partner” and officers will manage teams of AI agents in a hybrid model to increase the speed and scale of intelligence work.

Last year, the agency had more than 300 AI projects, and, for the first time in its history, used AI used to generate an intelligence report, he said.

Ellis' remarks provide a rare public glimpse into how the spy agency is integrating frontier AI systems into its day-to-day operations, and the expectation that they will soon become part of officers’ workflows.

The CIA primarily executes and coordinates human intelligence-gathering overseas, often undercover. Officers recruit and manage foreign assets to clandestinely gather intelligence in areas such as economics, terrorism, and cyber threats.

Much of that work involves the use of technology, though some argue that advanced AI tools may reinvigorate old-school tradecraft techniques.

But there have been benefits to technological investments. The agency recently elevated its Center for Cyber Intelligence into an entire mission center, a move that’s “paying dividends already by allowing us to deploy new tools to the field and gain more access to priority targets,” Ellis said.

“The battle of cybersecurity will be a battle of artificial intelligence,” and whoever best harnesses AI models will wield “enormous power,” he said. “Having a new mission center centered around cyber intelligence will put us on the path to secure the upper hand.”

The agency also recently announced a new acquisition framework to better integrate technology.

Ellis said the CIA doubled its technology-related foreign intelligence reporting, to track how foreign adversaries like China are using advanced AI and other technologies. Those intelligence products focus on technology use abroad and can include findings on areas like semiconductors, cloud computing, infrastructure, cybersecurity or R&D.

Ellis did not mention Anthropic’s Project Glasswing, a consortium announced earlier this week meant to help secure critical software against AI-driven attacks. The project was fueled by a powerful, non-public Anthropic frontier model the company says has already uncovered thousands of vulnerabilities but could be weaponized in the wrong hands.

The intelligence community and its industry suppliers are already examining and discussing how such a model may impact the future of cyber missions, Nextgov/FCW reported Wednesday.

Earlier this year, Anthropic declined to ease restrictions that kept its tools from being used for domestic surveillance or fully autonomous weapons. That led the Defense Secretary Pete Hegseth to designate the company's products as a “supply chain risk” designation and President Trump to order that all federal agencies phase out their uses of Anthropic tools. The company has legally challenged the move.

Ellis did not single out Anthropic, but he cautioned that the CIA “cannot allow the whims of a single company” to constrain its use of AI and said the agency is looking to diversify across multiple vendors to preserve operational freedom.

]]>

‘Hybrid constellations’ are making it hard for militaries to hide

Patrick Tucker — Thu, 09 Apr 2026 06:00:00 -0400

A planned satellite constellation will be able to image any location on Earth every 15 minutes and take more detailed images, a novel capability that could reveal even the nimblest and stealthiest military maneuvers, its developer says.

On Thursday, Vantor announced plans to enlarge its current fleet of 10 satellites “five-fold” with spacecraft that will produce images with 20cm resolution—better than its current 30- and 40cm imagery. The company also plans to add two dozen high-revisit/lower-res satellites. When the constellation is complete some time after 2029, Vantor officials said it will vault the company to the forefront of the space-imagery industry.

“The geo accuracy of our exquisite data combined with the revisit data—we can actually fuse that data and have highly accurate imagery that nobody else can do,” Vantor CEO Dan Smoot said in an interview.

The space-imaging race

The planned constellation is part of the increasing competition between space-imaging giants.

Maxar, which renamed itself Vantor in 2025, was created by the merger of Maxar Intelligence, with its two synthetic aperture radar satellites, and DigitalGlobe, whose Quickbird had a high-for-its-day resolution of 2.4 meters. The new company focused on satellites of higher and higher resolution.

Its services were generally complementary to those of Planet, whose cheaper, smaller, lower-orbit satellites collected lower-resolution imagery more often and over a much larger portion of the Earth. But in 2023, Planet became the first company to offer high-resolution imagery and persistent coverage when it launched the first of its six 30cm Pelican satellites.

Today, Vantor’s Smoot says the U.S. government and other customers are increasingly seeking imagery for damage assessment.

“We've seen a lot in Ukraine, of course, and we're starting to see that utilization in the current conflict” in the Middle East, he said.

Vantor’s planned Pulse constellation will enable it to take twice as many photos of a given area than its competitors, while its Vantage satellites will take pictures with 10cm finer resolution than Planet’s Pelican can produce. That sort of capability can reveal operations that are extremely difficult to spot, such as a submarine surfacing, as Vantor demonstrated in a picture it provided to Defense One.

But advantages in resolution and revisit are fleeting. Planet has its own plans to add higher- and lower-resolution satellites, the company said this week. And other competitors are emerging such as Austin-based SkyFi, which NATO selected for an accelerator program in December. More importantly, satellite images, once an exclusive capability of the U.S. military, are empowering adversaries. Russia is reportedly providing Iran with satellite imagery to help target U.S. forces. Other militaries want imagery to help guide strikes in GPS-denied areas or by artillery and rockets that lack internal guidance systems.

“As you can imagine, you need to be accurate within a couple meters,” said Smoot. NATO allies and other countries are “starting to really learn the value of that as they start to rearm, and they also start to realize that they don't have those inherent capabilities built into their national defense systems.”

But the rise of AI may be the biggest reason that Vantor and others are building hybrid constellations of high- and low-resolution satellites. A single system for high-res and low-res imagery and AI analysis allows customers to better manage the security of their data. The company’s Tensorglobe platform enables customers to obtain data without switching providers.

All of that is enabling a new era of satellite imaging that would give smaller countries the ability to target missiles as effectively as larger ones.

“That is actually one of the most unique propositions about us doing this hybrid constellation,” Smoot said. “The way we're doing it, it actually will bring that what we call targeting-grade categorization to sovereign nations.”

In other words, as the competition intensifies, hiding from the sky will become more difficult for everyone.

]]>

Army operations center is trying to solve battlefield data problems in real time

Meghann Myers — Tue, 07 Apr 2026 17:54:18 -0400

As the Army works to gather and organize data to support battlefield decisions, it has created a task force to help with small, short-term problems—and in the longer term, to shape the service’s overall approach to data management.

The Army Data Operations Center went live on April 3, service officials told reporters on Tuesday, and so far its small team of civilian and soldier data and software engineers have received seven requests from different organizations to help deconflict.

“It used to be about firepower, but it isn't really about that anymore,” said Lt. Gen. Jeth Rey, the Army’s chief of staff for command, control, communications, cyber operations, and network architecture. “It's really about who can get the data to make decisions faster, to dominate.”

The task force might help, say, to get a partner force’s data flowing into the Army’s next-generation command-and-control platform so that a U.S. commander can compare what the two militaries are seeing on one screen.

The ADOC is organized into a “warfighter engagement cell” that that triages requests, then feeds them to data engineers at the “finish cell” to come up with a solution, who then runs that by the “data management cell” to figure out what kinds of policies need to be created or modified to fix the issue in the long-term.

“Those things are actually a lot more difficult than what you think, to be able to do—because it might be different cloud environments, it might be different [areas of operations], data owners— and you need to go through all the access requests,” said Brig. Gen. Michael kaloostian, who heads the Command and Control Future Capability Directorate at Army Transformation

and Training Command.

ADOC has so far been fielding requests from units in training environments, he said, but it’s technically open to responding to troops in combat, and will prioritize those tickets.

“We haven't received anything yet to support those operations, but if there were to be a request, we would surge on that and prioritize that appropriately,” he said.

For the first 180 days, ADOC will respond to requests and track trends to give the Army feedback on which fixes can be incorporated into training or standard operation and whether the help-desk model is necessarily in the longer term.

“The Army will make informed decisions about what the structure should be and whether a centralized capability in the future is even needed, right?” Kaloostian said. “We just aren't mature enough as an Army right now to really, truly become data-centric. We need something that can aid the continuous transition and transformation of the Army to a data-centric force.”

It’s possible that the future looks like a centralized operations center that deals with the “higher-level heavy lifting” of organizing different data formats from their respective systems, so that soldiers on the ground aren’t having to sort through it themselves, said Lt. Gen. Chris Eubank, who heads U.S. Army Cyber Command.

“So I think the cyberspace domain is evolving at such a rapid pace that the organizations involved in that domain must evolve as well, but I think it's a necessary thing right now,” he added. “And the hope is we're creating soldiers that are data-smart more and more, and the heavier lifting is done inside of a central organization, if need be.”

]]>

As aircraft losses mount, Pentagon wants a software fix to see through the fog of war

Patrick Tucker — Tue, 07 Apr 2026 04:40:29 -0400

The U.S. planes that have gone down in the Middle East since the launch of Operation Epic Fury all lacked the same thing: a common operating picture that includes relevant intelligence and data. The Defense Innovation Unit announced Monday it’s looking for a fix.

The request: an open-architecture software suite to fuse real-time data into a credible picture of moving objects, threats, and conditions. The idea is to give pilots a broader understanding of reality in a way that is unremarkable to American motorists with easy access to data about ever-changing conditions but aspirational for air crews flying planes outfitted with antique computer hardware.

U.S. Air Force officials have warned of a lack of a common operating picture among airframes, particularly transport planes like C-130s, for years. But the loss of seven aircraft in just over a month, due mostly to communication errors and friendly fire, has exposed a big gap in how U.S. planes communicate with one another and with ground forces.

At the outset of the war on Iran, the U.S. lost three F-15E Strike Eagles due to Kuwaiti friendly fire. In March, they lost a KC-130 refueling tanker when it was involved in a mid-air mishap with a second tanker, due in part to a transponder failure but, again, pointing to a gap in the planes’ ability to fuse data to identify one another.

During the rescue effort that followed an Iranian shootdown of an F-15E Strike Eagle and an A-10 Thunderbolt on April 3, the U.S. destroyed two MC-130J transport planes when the planes were unable to take off from their makeshift runway, U.S. President Donald Trump said in a press conference on Monday.

“It was sandy, wet sand, so we thought there may a problem taking off because of the weight of the plane… And then we also had all the men jumping back onto the planes, and they got pretty well bogged down.”

It’s the sort of problem that access to real-time data on terrain, weather, and other factors could have solved. But most older transport planes lack up-to-date maps or terrain data, forcing crews to “rely heavily on pre-mission planning products, voice updates, and aging platform-specific displays,” according to DIU’s ask.

Because computer hardware varies widely throughout the U.S. aircraft fleet, the Air Force and aircrews frequently use workarounds such as software-defined radios or off-the-shelf communications equipment to get the data they need. But there is no common standard, which makes it difficult for aircrews to know what data they need.

“This problem is especially relevant for large, high-value airlift and tanker aircraft that utilize avionics and mission systems that are optimized for more permissive operations,” according to the announcement for DIU’s “Open Mission Engine” program. The new effort seeks software that can combine all the relevant friend, foe, intelligence, and logistics data into one place in real time, not afterward.

While the solicitation doesn’t mention Operation Epic Fury by name, the rising number of U.S. military aircraft mishaps shows how urgently the U.S. military needs a way to better let planes communicate with each other.

Among the features that DIU wants the new software “engine” to have is a “moving map” application that “uses relevant operational data into a single aircrew display, including blue-force awareness, threat and airspace overlays, mission updates, and route decision support.”

]]>

From launch to recovery: Here’s how Space Force is backing NASA’s Artemis II mission

Thomas Novelly — Wed, 01 Apr 2026 17:31:35 -0400

Space Force guardians are supporting NASA’s first crewed launch to the moon in more than 50 years by forecasting the weather and monitoring security, and, in an emergency, would make the call to abort Wednesday night’s historic mission.

The service’s Space Launch Delta 45 is tracking by weather and rocket safety conditions for NASA’s launch out of Florida’s Kennedy Space Center. There will be 28 crew members in the mission control center, in contrast to the four or five members used for a typical launch, service officials told reporters during a roundtable Monday.

Boeing’s Space Launch System rocket doesn’t have an autonomous flight safety system, which would automatically stop the mission in an emergency. That means more eyes are needed to monitor key data on the rocket and in the skies.

“When you don't have an autonomous flight safety system, you have to call up multiple range assets and resources for that function,” said Lt. Col. Gregory Allen, 1st Range Operations Squadron commander. “So that is exactly why we have so many personnel on console for this.”

The last Apollo mission in 1972 marked the last crewed mission to the moon. The Space Force, since it was created in 2019, has taken on a major role in rocket launches and oversees military and U.S. government missions from Florida’s Cape Canaveral and California's Vandenberg Space Force Base. Additionally, the 1st Air Force’s Detachment 3, the only Defense Department unit tasked with rescue and recovery of commercial space crews, has four helicopters ready at Patrick Space Force Base in case the crew needs to abort the mission.

Service officials told reporters that airmen and guardians are ready for Wednesday’s launch and opened an emergency operations center manned with first responders, bomb technicians, and others they can “surge” if needed, said Col. Chris Bulson, the deputy commander for installation support.

“SLD 45 has a robust team of folks on standby for this particular mission that are well versed with the SLS rocket, the systems on board, and have great teamwork with our detachment three partners to assist them should a recovery be necessary,” Bulson said.

Given the large public crowds for the event, Space Launch Delta 45 is also preparing for more security patrols and more people to manage crowds on land as well as boaters watching from the ocean, he added.

The Delta’s 45th Weather Squadron has forecasted favorable conditions for Wednesday’s launch. Unlike the National Weather Service or the National Oceanic and Atmospheric Administration, the Space Force’s squadron uses guardians who focus on specific factors and local weather that would affect crews, missions, and launch vehicles and pads.

They also monitor last-minute winds and conditions in case the Artemis II crew needs to abort mid-flight, Col. Douglas Oltmer, the weather squadron commander, told reporters.

“If the winds are too strong on shore, and they do have to abort en route, they will need to land over the water for the emergency,” Oltmer said. “So that's one of the things we're forecasting for on our end as well.”

The squadron’s weather forecasting doesn’t stop once the crews enter outer space. The Space Force has a team alongside the 1st Air Force’s Detachment 3 aboard the Navy’s USS John Murtha transport that will monitor recovery conditions for the crew when they splash down in the Pacific Ocean after their 10-day journey. The four astronauts are scheduled to launch from Kennedy Space Center sometime after 6:24 p.m. on Wednesday, and are scheduled to return off the coast of California on April 10.

“With Artemis, we started back in November, providing all the support all the way to day of launch, all the way until they recover, and then we actually forecast to bring the capsule all the way back across the country from the West Coast,” Oltmer said. “It's probably one of the coolest jobs an Air Force meteorologist can have here.”

]]>

AI may revive old-school tradecraft even as it transforms intelligence work

David DiMolfetta — Wed, 01 Apr 2026 09:00:00 -0400

Artificial intelligence is widely expected to revolutionize intelligence-gathering, enabling faster, cheaper and more scalable collection of information. But a new analysis suggests the technology may also spur a return to some of espionage’s oldest methods.

A recent article in Studies in Intelligence, the CIA-backed academic journal, argues that as AI degrades the reliability of digital communications like text messages and video calls, traditional human intelligence tradecraft — like dead drops, brush passes and in-person meetings — could regain renewed importance.

The same technologies that improve intelligence gathering may make it harder to trust the data those tools produce or transmit, argues the author, Thomas Mulligan, a RAND Corporation researcher who served in the CIA from 2008 to 2014.

AI is already being used to generate convincing deepfakes and fabricate messages. Mulligan argues that these introduce a new source of “noise” into digital communications, which makes it harder to distinguish between authentic and synthetic signals.

That has implications for how spies communicate with their sources.

“If my friend tells me, face-to-face, that he is in trouble and needs money, I can be confident that that’s true,” Mulligan writes. But when the same message is delivered through an electronic medium, it becomes “more likely a scam than a bona fide plea for help.”

That dynamic elevates the value of communication methods that are not mediated through electronic means.

A properly executed dead drop, for instance, allows an intelligence officer to securely receive information while also verifying that it came from a specific human source, rather than an AI-generated deception, he says. A dead drop involves a secret location used to exchange information or physical items between people without requiring them to meet face-to-face.

The same logic applies to brief, in-person exchanges like brush passes, in which spies and sources pass materials to one another during a quick, seemingly routine encounter in public.

The argument runs counter to assumptions that advances in AI will diminish the role of human intelligence, or HUMINT, in favor of more technical collection methods.

Long before the advent of spy satellites and tailored computer hacking kits, human intelligence dominated espionage as the world’s oldest form of spying. From royal couriers and informants in the Persian Empire carrying sensitive information across imperial networks to the Culper Spy Ring’s use of invisible ink and dead drops during the American Revolutionary War, intelligence once solely moved through people.

In recent months, the Trump administration has made it a point to highlight contributions that CIA operatives have made toward its national security achievements, including efforts targeting the government of ousted Venezuela leader Nicolás Maduro. The agency has also taken a more public-facing posture, releasing recruitment videos aimed at sourcing in China. And in the months leading up to the Iran war, agency spies had been reportedly tracking the movements of now deceased Ayatollah Ali Khamenei.

At any given time, the CIA, the nation’s primary human intelligence agency, may be operating across dozens of countries worldwide to collect foreign intelligence or conduct covert action — activities intended to influence political, economic or security conditions abroad, while concealing the U.S. government’s role.

Mulligan’s paper also comes as the tech industry has pushed for AI adoption across government agencies, including offices focused on national security and intelligence. In February, the CIA announced a major overhaul of its technology procurement process, as part of an effort to more quickly adopt leading-edge capabilities for use in its missions.

In a phone interview, Mulligan said AI may play a more permanent role in helping human spies craft better-sounding communications, just as cyber experts have argued that AI tools greatly enhance and scale bad actors’ phishing campaigns.

“A core part of being a case officer and human intelligence operations is persuasion, talking to a prospective agent or a recruited agent and trying to convince him or her to do things that can be difficult, can be dangerous and can be stressful,” he said. “I think AI has a constructive role to play, from the point of view of a case officer, in enhancing his or her ability to persuade.”

But there’s a prevailing question about how much intelligence practitioners risk when they outsource tasks to AI. Gathering intelligence from other people “is a human business at the end of the day, and it does involve an agent and a case officer as a team engaging in a difficult and sometimes dangerous relationship,” Mulligan said.

“My view,” he said, “is that [HUMINT] will have to have a human element — a real, essential human element — for the foreseeable future.”

]]>

Startup debuts agentic AI assistant for war

Patrick Tucker — Wed, 01 Apr 2026 06:00:00 -0400

The Pentagon is eager to incorporate AI “agents”—software that can autonomously execute complex tasks like customer service, scheduling, or code writing—into more of what soldiers and defense civilians do. But a growing body of research shows that agents built from well-known large language models exhibit unpredictable and dangerous behaviors even in benign settings.

Edgerunner AI, a veteran-founded startup, built a different kind of agent tool for the military, one trained by former operators and experts on actual military tasks and in real combat settings. The Wednesday release of the new tool, WarClaw, is part of a trend away from large, big-name frontier models toward smaller, more custom ones that offer more user control.

Public interest in what’s called “agentic AI” rose 6,100 percent between Oct. 2024 and October 2025. Demand for software that can autonomously achieve complex tasks by designing and implementing processes and then fine-tuning the results without continuous prompting is forecast to rise from $4 billion last year to more than $100 billion by 2030.

The Defense Department is moving on the trend early. In January, as part of its AI strategy rollout, it announced development of an “Agent Network” to build “AI-enabled battle management and decision support, from campaign planning to kill chain execution” and to build a “playbook for rapid and secure AI agent development and deployment” for business processes.

Edgerunner’s AI agent, WarClaw, “searches and analyzes databases, interprets intelligence reports, pulls relevant information from the web, drafts documents and briefings, and automates routine processes. Integrations include Microsoft PowerPoint, Word, Excel, Teams, Outlook, and more,” according to a company statement. But it is very different from similar tools from well-known model makers like Anthropic, xAI, or OpenAI.

Tyler Xuan Saltsman, the founder of Edgerunner AI, told Defense One in an exclusive interview that agents that come from such companies pose a particular risk to the military; the claim is backed by recent scholarship.

In March, scientists from Harvard, MIT, and others found agents built from Anthropic’s Claude or Kimi, and then run in OpenClaw (agent software that works with large language models), exhibited “unauthorized compliance with non-owners, disclosure of sensitive information, execution of destructive system-level actions, denial-of-service conditions, uncontrolled resource consumption, identity spoofing vulnerabilities, cross-agent propagation of unsafe practices, and partial system takeover.”

AI agents also offer an “illusion of control,” which is particularly dangerous in military contexts, according to a March paper from Cornell University. The researchers found that agentic systems can absorb corrections or resist assessments in ways that military planners and monitors can’t see because the processes to expose them don’t exist. “A waypoint-following drone cannot misinterpret an instruction; a pre-programmed targeting system cannot absorb a correction; a conventional sensor network cannot resist an operator’s assessment. Agentic systems can do all of these things, and current governance frameworks have no mechanisms for detecting, measuring, or responding to these failures,” the authors write.

AI agents derived from the most well-known large language models also don’t like following orders from military commanders, rejecting commands some 98 percent of the time, according to a paper Saltsman co-wrote. He told Defense One that his time working with these models compelled him to adopt a completely alternative approach to model development.

WarClaw

Products from OpenAI, Anthropic, and Grok represent only one way to develop high-functioning models—a method built on harvesting huge amounts of data from the internet and then both training and running them in large, energy-intensive data centers. Because these models are mostly consumer-facing, they’re designed to keep users asking questions, contributing prompts and data, and looking at advertisements. Those incentives help to explain why the models respond to users with flattery and reassurance, even when users are wrong.

Saltsman said the chronic inclination toward sycophancy in popular large language models is a national security problem, broadly. But for the military, it’s an even bigger risk.

So Edgerunner AI took a different tack. They use large enterprise cloud resources to train the models, but the models are capable of running on premises with no internet connection, which is essential for military operations. That gives the user much more control over how much (or little) energy goes into them. Unlike better-known large language models, they are trained on a highly curated data set that’s specific to the military, and the trainers include military subject matter experts and former operators.

Most importantly, he said, the agents are designed to run autonomously, saving the operator time and attention. But autonomous does not mean without human supervision or control. The models can’t just pick whatever strategy they might like to complete a task without their operator's permission. And the processes are designed to be auditable and transparent, as opposed to the opaque functioning of other models.

That unique approach has caught the attention of military users via contracts and cooperative research and development agreements with the Kennedy Special Warfare Center and School that trains special forces groups, and Special Operations Command. The company is working with the Navy to integrate their software onto submarines and warships, via the Interagency Intelligence and Cyber Operations Network, and they are working with Lockheed Martin and the Army on the Next Generation Command and Control system.

The unique attributes of their approach to model building—curated data, communication independence, and control over process—are also what civilian users increasingly want from AI, according to surveys. That suggests future dual-use potential for the company in an environment where both consumers and warfighters are looking for alternative futures for AI.

]]>

The Army wants to use bullets, mortars, and artillery to take out small drones

Meghann Myers — Thu, 26 Mar 2026 03:41:31 -0400

HUNTSVILLE, Alabama—Army leaders often say they want to stop using million-dollar missiles to shoot down thousand-dollar drones. One of the solutions may be cheap munitions the service already has.

While the Army is leading an interagency task force to source dedicated counter-drone systems, the Capability Program Executive for ammunition and energetics is taking a look at the Army’s existing rounds to make them capable of shooting down much smaller targets.

“We all know how important missiles are to the fight. We see it in the news. But there's also a point where missiles can't get after every munition, every threat, so we need to supplement that with something that we already have within our formations,” Kaitlyn Tani, deputy project manager at Maneuver Ammunition systems, said Wednesday at the AUSA Global Force Symposium.

There has already been demonstrated success in using bullets against drones, she said, particularly ones like the XM121 High Explosive Proximity round, a 30mm bullet that only needs to get near a target to take it out with its blast radius.

There are a handful of other systems that can use this proximity fuze technology and put it into weapon systems that aren’t made to shoot tiny quadcopters in the sky, but could do it with a round that only needs to get close.

“We're taking your Bradley fighting vehicle and [making it] counter-UAS capable by using the armament that is already on the system,” Tani said. “We're taking the infantry soldiers who already have Mk-19s within their squad and providing them with counter-UAS capability.”

There’s also the potential to take some larger rounds, like indirect fire munitions designed to be launched at coordinates rather than targeted using the sight of a gun, and strapping them onto drones.

“So we're definitely looking at all of our legacy munitions,” said Col. Vinson Morris, project manager for Close Combat Systems.

To do that, Morris said, they have to tweak the rounds to use them with a UAS, as they’re designed to be fired from a ground-based launcher that does the propelling toward a target.

“We see the first iteration of this potentially being a legacy mortar or legacy artillery [round] dropped from a UAS,” he said.

]]>

The real danger of military AI isn’t killer robots; it’s worse human judgement

Patrick Tucker — Wed, 25 Mar 2026 21:48:33 -0400

The debate about lethal autonomy—core to the Trump administration’s fight with Anthropic—obscures a deeper danger of the Pentagon’s rapid adoption of commercial AI tools: they might weaken the U.S. military’s ability to tell fact from fiction.

New research suggests that relying on AI to do various tasks can erode one's native ability to do them. Military commanders are taking note.

“The more you use AI, the more you will use your brain in a different way,” said French Adm. Pierre Vandier, NATO Supreme Allied Commander for Transformation, in a conversation in Defense One’s "State of Defense" series. “And so [we need to] be able to have some oversight, to be able to critique what we see from AI, and to be sure you are not fooled by a sort of false presentation of things. It's something we need to take care of.”

But there is scant evidence that the Pentagon, in its rush to deploy AI tools, is taking steps to keep its users sharp—or even to monitor the effects of AI.

“Because of the pace and the urgency associated with deploying these models, maybe that hasn’t been put in place,” said one former senior military official who worked to deploy AI in combat settings.

In the meantime, the pressure to use AI tools is only growing.

“Especially as you get deeper into any conflict, there will be more and more pressure to find more targets. That happens in every conflict as well. After two weeks, you're sort of going through your delivery target list. Now you're demanding targets. Well, how fast can you generate targets?” the former senior military official said.

A cognitive trap

A growing body of research shows that wide use of large language models can undermine human thought and communication.

For example, it can homogenize thinking among users, reinforcing “dominant styles while marginalizing alternative voices and reasoning strategies,” according to an Air Force Research Laboratory paper published earlier this month in the journal Cell.

This presents at least two problems for the military, said Morteza Dehghani, a University of Southern California computer science professor who helped write the paper. In the moment, “it washes away signals about who the author is,” and therefore eliminates important context for evaluating data. And over time, it can stifle critical thinking.

“Because these models are optimized for the most likely and ‘idealized’ responses, they often enforce a linear, ‘Chain-of-Thought’ reasoning style,” Dehghani wrote in an email. “This can disincentivize experienced analysts from employing the non-linear, intuitive, or ‘gut feeling’ strategies that are essential for identifying rare exceptions or navigating complex, non-standard intelligence scenarios.”

Similarly, researchers at Wharton found in January that people using LLMs spend less and less time scrutinizing results for accuracy or employing their own judgment. They found that users rely on the AI’s judgement even when they know it’s wrong, a phenomenon the authors call “cognitive surrender.”

And a February paper from Princeton found that the way LLMs speak to users—referred to as “sycophantic AI”—instills a false sense of confidence and isolates people within preconceived biases: “Our results show that the default interactions of a popular chatbot resemble the effects of providing people with confirmatory evidence, increasing confidence but bringing them no closer to the truth.”

Who is driving?

The military is aware of at least some dimensions of the problem. Speaking on a March 6 podcast, Pentagon research-and-engineering chief Emil Michael said that his core concern about Anthropic—which President Trump has barred from use by the federal government—was that military users might become too dependent on a single untrustworthy tool.

“Maybe it's a rogue developer who could poison the model to make it not do what you want at the time, or sort of trick you because you have to trick it. I mean, all these things that we know when we worry about models that hallucinate purposefully or do not follow instructions,” Michael said.

Anthropic had similar concerns: that its tools might be used by people untrained to properly evaluate its output. One company official said their chief worry was that they had not validated that the model could be used reliably for compiling targeting lists. Worse, they had no way of knowing how the military was putting their tools to use. For example, Anthropic officials only learned after the fact that their tools had been used to plan the Jan. 3 raid into Venezuela.

The Pentagon has since declared Anthropic products a supply-chain risk and is working to replace them, though they are still in use by military planners, including in the Middle East.

A former senior official described it as “a serious governance question.” Frontier AI companies like OpenAI, Anthropic, Google, and xAI cannot provide the in-depth support required for military units to understand the conditions under which these tools are used.

“It's almost becoming a journey of discovery for the government,” the former senior military official said. “Are there people on site from these companies helping the day-to-day user? My guess is, if there are, there may be only one or two of them.”

]]>

Defense One Radio, Ep. 205: New heat science and the future of soldiering

Defense One Staff — Wed, 11 Mar 2026 19:48:41 -0400

Apple Podcasts

Guest:

Lt. Col. David deGroot, director of the U.S. Army Heat Center at Fort Benning, Ga.

See also Patrick Tucker's related recent reporting on the topic, here.

]]>

New science on heat is changing the future of soldiering

Patrick Tucker — Mon, 09 Mar 2026 18:52:56 -0400

Wearable biometrics, improved science, and more data are changing U.S. Army attitudes toward human performance—particularly how soldiers adapt to the risks of overheating.

The U.S. military has been studying the effects of heat on troops for almost a century, dating to the 1927 establishment of the Harvard Fatigue Laboratory at the military's request. Still, soldiers’ and commanders’ approach to core physical tasks—think timed runs, strenuous outdoor activity, or environmental exposure—lags the growing body of science about heat risks, sometimes by years or decades. That may finally be changing under new initiatives to expand research into human performance.

Lt. Col. David deGroot, who directs the Army Heat Center, told Defense One that new wearable devices—which track metrics like heart rate and body temperature—are providing military trainers and soldiers with a new window into how their bodies interact with the environment. This data offers insights well before physical symptoms of overheating appear. That timing is key: by the time a soldier experiences heat exhaustion or the first symptoms of a far more severe heat stroke, it is often too late to prevent a health incident that could sideline them from training or worse.

“We've got some still-evolving, non-invasive, wearable technologies that are going to be able to, not necessarily prevent a heat stroke, but detect it much, much sooner”—as many as 12 minutes sooner, giving time to reduce its severity, deGroot said on the sidelines of the March 4 Heat Forum at Fort Benning, Georgia. (Disclosure: the event was staged by FBC, a Defense One sister brand.)

New data also reveals the costs to units and individuals when they push too hard under the assumption that suffering makes a soldier tougher. The idea that ignoring discomfort is the hallmark of "good soldiering" goes back centuries, but deGroot argues it needs an update.

Too often, he said, it is the soldiers themselves, particularly those trying to qualify for Ranger School, who exhibit a specific risk factor for heat-related illness: a trait that might look good on a resume as being “highly motivated.” Troops who push themselves to attain a new personal best in every physical task run a much higher risk of suffering a heat stroke and taking themselves out of commission.

DeGroot is on a mission to remind soldiers, trainers, and commanders that it doesn’t make sense to attempt a personal record during every march or run; sometimes, it is best to simply hit the acceptable threshold and save the record-breaking for another day. He also clarified that simply drinking water is not a shield against heat sickness. In fact, he noted, 80 percent of heat stroke victims are actually well-hydrated.

“If you successfully hydrate, congratulations: you have successfully mitigated the risk of dehydration. But if you are sick, if you are not acclimatized, if you are overweight, or if you are excessively motivated, those risks still exist despite appropriate hydration,” he said.

He added that trainers too often confuse periods of physical rest with "inactivity" and grumble about science-backed ratio tables that dictate appropriate activity levels based on external factors.

“When you look at those tables, they're not as restrictive as you think they might be,” he said. “There are a lot of things you can still be doing that are productive training or operational activities.”

The new science on heat isn’t just slowing training down. In some ways, it is speeding it up. Take research on acclimatization—the process of adjusting to different temperatures. Previous thinking suggested that soldiers needed as long as eight days to adjust to a new climate. However, research published last year showed a different approach. Researchers compared a standardized eight-day acclimatization group (90 minutes a day) against a group that performed four 90-minute bouts in one day and four more the next. Both groups reached the same level of acclimatization.

“When I read it, I was excited about it," deGroot said. "This is a fascinating idea that I never even thought of.”

]]>

Meet the startups trying to build military-specific AI

Patrick Tucker — Sun, 08 Mar 2026 08:00:00 -0400

The battle between AI model builder Anthropic and the Pentagon has exposed a huge gap between what AI tools the military wants and what companies like Anthropic, xAI, and OpenAI actually make: AI tools for use by everyone, not specifically for the military. A handful of veteran-run or -financed startups aim to fill that gap. Their pitch: AI for war should have some basic understanding of war, beyond reading Tom Clancy fan fiction. It shouldn’t confidently offer low-confidence answers just to appease the user. And it should work even when a high-tech adversary severs its connection to the cloud.

The needs gap

Among the uncomfortable truths the fight between Anthropic and the Defense Department reveals is that the Pentagon had deep reservations about the language models themselves, their potential for hallucination, and that they may “not follow instructions.”

But the Pentagon allowed wide deployment of Anthropic’s model anyway, anxious to get at least some generative-AI tools into operators’ hands. It reportedly played a role in Operation Midnight Hammer, the raid that captured Venezuelan President Nicolás Maduro, although Pentagon officials have declined to confirm that.

After the raid, Anthropic officials called Palantir to ask whether their AI models had been used in the operation, Defense Undersecretary for Research and Engineering Emil Michael said on Friday. Michael said that was “a whoa moment for the whole leadership at the Pentagon, that we're potentially so dependent on a software provider without another alternative.” He said it raised several concerns, including that Anthropic might shut down access to models in such situations.

Anthropic itself had similar concerns, according to one company official: the company didn’t think it was safe for the military to rely on their models for combat situations.

Another aspect of the shortcomings of today’s frontier AI models—Anthropic’s Claude, Google’s Gemini, OpenAI’s ChatGPT, and xAI’s Grok—is that they need a connection to the cloud. This makes them unreliable for today’s troops and unusable for tomorrow’s autonomous weapons.

OpenAI tacitly acknowledged this limitation when it recently announced its own deal to deploy on the Pentagon’s classified networks—though it described this inability to deploy large foundational models to the battlefield as a “safeguard” against the kind of unreliability that concerned Anthropic officials.

“Our contract limits our deployment to cloud API,” OpenAI’s national security lead Katrina Mulligan explained on X. “Autonomous systems require inference at the edge. By limiting our deployment to cloud API, we can ensure that our models cannot be integrated directly into weapons systems, sensors, or other operational hardware.”

The way forward

Even as the Pentagon was noisily ejecting Anthropic from its good graces, the Army was preparing to unveil a new effort to close the gap. Project Aria, announced on Thursday, is intended to help the service develop and deploy new AI models and tools “to tackle real operational problems”—that is, designed specifically to help soldiers do their jobs.

This is also the object of a new class of AI startups run by people with military experience and dedicated to battlefield tools that don’t need to phone home.

One is Smack Technologies, which on Monday announced that it had secured $32 million in investor funding to build what they call a “frontier lab for national security.”

Andrew Markoff, a former Marine special operator who co-founded Smack, says his AI is trained on combat-relevant datasets, not the unspecialized fodder fed to Claude, Gemini, and other frontier models.

“There is no training set for World War Three, right?” Markoff said in a call with reporters last week. “There's no way to build reinforcement learning…if you don't have deep domain expertise and a deep bench of people with domain expertise. There is no shortcut around encoding good human prior knowledge, and it doesn't exist in doctrinal manuals.”

He called the Venezuela raid a good example of the sort of operation that AI could help scale up in a conflict with a more advanced adversary.

“Multiply it by 100 and scale. You have targets that you want to strike, you have sensors that you're trying to allocate on those targets to figure out what's going on. And to facilitate the strikes, you have strike platforms and escorts that are coming together from all over the world with very detailed sequencing requirements; you know, task A has to happen X number of seconds before Task B. And all of these are dependent on some, some other thing happening at, you know, time X. So, like, all of these things have to come together globally, a really tight timeline.”

But Markoff said that’s not the sort of thing that commercial large language models are built to do. Models like Claude “have no way to optimize between those goals. And it doesn't have the ability to do the detailed time, space calculations, [to perform] geospatial reasoning grounded in physics, to make the decisions about, literally, which munitions need to be where at what time, talking to which sensors at what time. It doesn't have the ability to do that.”

This was echoed by Jason Rathje, a former Air Force acquisitions officer and co-founder of the Defense Department’s Office of Strategic Capital who now leads the public-sector division at webAI.

Frontier models like Claude “are built to answer millions of different kinds of questions for billions of users. Military organizations often need something different, systems tuned for specific operational tasks like logistics planning, equipment maintenance, intelligence analysis, or operational decision support,” Rathje said.

The limitations related to cloud needs are equally important. “Many of today’s frontier models are designed as centralized services for massive commercial user bases, requiring the most advanced chipsets and high-capacity data center infrastructure available, and consuming enormous amounts of power. That makes sense for consumer applications, but military organizations often have very different requirements,” he said. “What defense organizations are asking for is sovereignty: control over the model, the data, and the infrastructure it runs on.”

Smack Technologies is producing two product suites: one to work like the well-known generative AI models, but trained on military intelligence and operator experience; and the other to work in remote battlefields.

Sherman Williams, a Navy veteran and founder of AIN Ventures, has invested in a number of dual-use and defense-focused startups. He acknowledges that no AI startup is going to beat one of the big frontier models in metrics like reasoning benchmarks. But “a model that's 85% as capable but runs on a [denied, disrupted, intermittent, and limited] network at the tactical edge beats GPT-5 in a data center you can't reach.”

Even the data centers you can reach are vulnerable, as shown by Iran's targeting of an AWS facility in Bahrain. “These data centers are important, but they are also vulnerable. Context matters more than benchmarks.”

The new class of DOD-focused AI startups “aren't trying to out-train OpenAI,” he said. “They're building the adaptation and deployment stack that makes open-source models usable in classified settings. Secure fine-tuning, domain-specific models for [intelligence, surveillance, and reconnaissance] and [command and control] edge deployment.”

Williams says he’s seeing “strong pull signals from military customers, especially SOCOM and INDOPACOM,” which has been extensively using AI at a headquarters level for more than a year.

He added that DOD buyers and users want to trust the makers of their AI tools, and that bond is easier forged with founders who are familiar with military operations.

But just hiring a veteran-developed AI doesn’t solve a broader problem of large language models: they speak confidently when they shouldn’t, and they often tailor their responses, or even lie, to their users.

Pete Walker, a retired Navy commander and the chief innovation officer at defense AI and cybersecurity firm IntelliGenesis, said that the big frontier models often provide answers users want to hear.

“The way these models are built, one of the reasons why they're so big, is they encourage conversation,” and that means encouraging users to dive deeper into areas of interest on specific topics, not talking to them honestly. Walker, who holds a Ph.D. in cognitive science, has peer-reviewed research to back up these assertions.

So his company is working to develop a framework for large language models based on counter-factual thinking—presenting alternative points of view to challenge users' assumptions rather than simply reinforce the assumptions the user brought to the original question. He describes it as getting a model to think, ‘Hey, you're saying that if A then B, but what if it's not A, or what if not B? What does that imply?’ And so I think those are areas of research that we need.”

]]>

25th ID helping the Army smooth out the wrinkles in its next-generation C2

Meghann Myers — Thu, 26 Feb 2026 15:29:38 -0500

The Army wanted a command-and-control platform that would sync up its many siloed battlefield data management systems—from intelligence to surveillance to targets to munitions levels—and display it all in one place. Once they had it, soldiers realized they needed a little more control over how much information sat on their screens at one time, so the developers got to work on a fix for that.

This is one example of the way the Transformation-in-Contact model is helping to shape the Army's Next Generation Command and Control effort, or NGC2, courtesy of the Hawaii-based 25th Infantry Division and their Lightning Surge series of exercises, the second of which is underway now following a first iteration in January.

“It was somewhere in the thousands of data objects…there wasn't a way to control the flow from Palantir. You just either got it all or got nothing,” Lt. Col. Adam Brinkman, 25th ID’s head of communications and network, told reporters Wednesday. “So in between Lightning Surge 1 and 2, there was a really, really good link up between Lockheed Martin and Palantir to develop an application within Palantir that allowed us to select specific things and push those to the data layer, as we really needed.”

Coming out of Lightning Surge 2, he added, there will need to be an application that separates and organizes unclassified versus classified data as it gets fed into the system.

“Eliminating that swivel chair is really the key objective that we've got to get down,” said Maj. Gen. John Bartholomees, 25th ID’s commander, in a nod to switching back and forth between classified and unclassified computer systems at a desk. “We communicate often and well with the joint force, but it takes hours and energy that should be automated at this point, that we're having to do manually.”

Speaking of automation, 25th ID is also looking for an automatic way for NGC2 to select which type of satellite it’s using to sync up systems, so soldiers don’t have to manually switch based on whether public or private 5G, for example, has the best connectivity at that moment.

“How to make sure that transport path is going over the best possible route to synchronize data across the entire division,” Brinkman said. “We've done some manual changes through Lightning Surge 2. We were able to make it work…just every time we've done that so far, it's been a significant engineering effort to bring that platform asset back up as we're switching between those transport options. So we definitely need to speed that up. It needs to be self-sensing and then self-determining.”

All of these tweaks and upgrades are meant to push NGC2 to a platform that helps commanders make decisions quickly enough to get ahead of an adversary’s next move.

“One specific in this—we think this is a very achievable goal—is essentially from a time of electronic warfare sensing of an adversary to a round impacting is less than four minutes,” Bartholomees said. “And that's with an unknown adversary with an unknown location and with a single target.”

The division will take the new updates out for another spin during Lightning Sugar 3 in April.

]]>

Air Force test pilots used tactical AI to evade a missile

Thomas Novelly — Tue, 24 Feb 2026 16:29:22 -0500

AURORA, Colorado—Air Force test pilots used artificial intelligence aboard an experimental fighter jet to successfully evade a simulated incoming missile, showcasing how the service’s aviators may rely on AI in a future fight.

Lockheed’s secretive Skunk Works research arm acknowledged the experiment Monday during the Air and Space Forces Association’s conference here. Late last year, test pilots at Edwards Air Force Base, California, received a simulated warning for an incoming surface-to-air missile while flying Lockheed’s experimental X-62A Vista jet. The onboard AI detected the missile and, without the pilot’s control, conducted an evasive maneuver.

“In this case, the missile signal or warning came in, the pilot didn't have to do anything, and the aircraft responded in a tactically appropriate way to keep the pilot alive and preserve the aircraft,” OJ Sanchez, Skunk Works’vice president and general manager, told reporters.

The test was called “Have Remy,” named for the rodent who helps a French chef cook by controlling his movements in the Disney film “Ratatouille.” The project also illustrates how AI tools might be used by the service’s fighter pilots. Distrust in AI among the general public still remains high, which experts have said may have broader national security implications.

Skunk Works’ project helped Air Force pilots train its AI models while simultaneously offering an opportunity for pilots to help develop and see how the technology may benefit them in future fights. Sanchez said the project showed how a fully autonomous unmanned aircraft could perform evasive maneuvers or be used as a feature in a suite of tools for aviators.

“Exercises like Have Remy are part of changing the way that we think about using AI agents in some of the most stressing human conditions, and I can't think of one more stressing than inside a fighter cockpit under attack.”

The X-62A Vista is a modified version of the F-16D Fighting Falcon, used to test automation and artificial intelligence. In 2024, Air Force Secretary Frank Kendall flew in an X-62A piloted by Shield AI’s software in a simulated dogfight with a manned F-16 fighter.

A September survey from the Pew Research Center detailed American skepticism and lack of trust in AI. Fifty percent of respondents said they’re more concerned than excited about the increased use of AI in daily life, and a majority of those surveyed also believe the technology will only make problem solving worse. Sanchez said experiments like “Have Remy” hope to break down those hesitations.

“We are going to have to get all of us as humans comfortable with working alongside artificial intelligence, and that takes trust, just like a human-to-human relationship,” Sanchez said.

]]>

An MQ-20 drone just teamed up with an F-22 for mock combat missions

Thomas Novelly — Mon, 23 Feb 2026 05:00:00 -0500

AURORA, Colorado—A General Atomics MQ-20 drone took orders from an F-22 pilot during a recent mock mission that demonstrated robot-wingman concepts, company officials said in a Monday statement.

During a test flight earlier this month at Edwards Air Force Base in California, the pilot used Autonodyne’s Bashi pilot-vehicle interface to order the autonomous drone to “execute tactical maneuvers,” move waypoints, conduct combat patrols, and take on “threat engagement tasks,” the statement said.

“We appreciate the flawless execution of this mission using the government’s advanced autonomous systems,” said GA-ASI President David R. Alexander. “This demo featured the integration of mission elements and the ability of autonomy to utilize onboard sensors to make independent decisions and execute commands from the F-22.”

The flight followed a November demo in which an F-22 pilot used a tablet to control an MQ-20 using L3Harris datalinks and software radios with Lockheed Martin’s open radio architectures.

General Atomics, Anduril, and Northrop Grumman are all in the running to build the Air Force’s first collaborative combat aircraft. Earlier this month, the service announced that it had used the government-owned Autonomy Government Reference Architecture, or A-GRA, to integrate RTX Collins software aboard General Atomics’ YFQ-42 CCA aircraft and Shield AI’s technology on Anduril's YFQ-44 CCA.

General Atomics has since said it had logged another semi-autonomous flight on its YFQ-42 drone wingman with RTX Collins’ autonomy software onboard. Anduril and Shield AI, as of last week, had not had a joint CCA flight together. Northrop plans a first flight for its drone wingman this year.

In a separate Monday statement, General Atomics said it had given the name “Dark Merlin” to the YFQ-42—a reference, it said, to deadly falcons and “the wizardry of Merlin from Arthurian legend.”

Anduril, whose company takes its name from a sword from J.R.R. Tolkien's “The Lord of The Rings” fantasy books, calls its CCA offering “Fury”—the original name given to the aircraft by Blue Force Technologies, which was acquired by Anduril in 2023. Northrop’s “Project Talon” CCA is reportedly a nod to the Air Force’s T-38 trainer.

“Dark merlins are hunting machines, built for speed and aerodynamics,” Alexander said in an emailed statement. “They harass other falcons for fun, and they eat what they kill. The name sums up our new uncrewed fighter perfectly.”

]]>

New Army office aims to quickly develop and scale soldier ideas

Meghann Myers — Fri, 20 Feb 2026 22:39:04 -0500

A new Army organization aims to meld service leaders’ mandates to cut bureaucracy and adopt a venture-capital mindset with their push to turn soldiers’ ideas into manufactured gear.

The Pathway for Innovation and Technology is another office, yes, but its director says it will bring together the Army’s innovation and rapid acquisition hubs at the service’s headquarters level, putting muscle behind what until now have been siloed efforts, and coordinating them with the Program Acquisition Executives, who have the funding and authority to turn ideas into programs.

“I sit at the table with the PAEs. I communicate directly with them to express what we are seeing. I also serve as a tech scout for them,” Col. Shermoan Daiyaan, PIT’s director, told reporters Friday. “What we've had is unit-driven innovation. We've had lab-driven innovation with [program managers] and PAEs. But in this case, the gloves are off and we can inject that capability.”

Within PIT sits the Army’s FUZE program, Joint Innovation Outpost and Global Tactical Edge Directorate, who through Daiyaan now have a direct line to the acquisition offices.

“What I think is really interesting here is these divisions and corps have had these innovation hubs or labs for a while now, and they would find ideas, but there was no path to go take it to scale,” said Chris Manning, the Army’s deputy assistant secretary for research and technology. “And so [members of the PIT organizations] are there on the ground and saying, ‘Hey, what are the most promising things coming out of there? And how do I connect this to the broader acquisition enterprise?’ ”

One of the top issues they’d like to tackle is power generation and integration, particularly for unmanned aerial systems, Daiyaan said.

“With us doing multiple vendors of UAS, multiple vendors of different capabilities, they all may have a different battery,” and it might be nice to have “one charger to charge them all,” he said.

Ideally, a PIT-lead program development might look like an xTech competition to get a bunch of companies working on a problem the Army wants to solve, select maybe 10 companies and award them prize money, get their prototypes in the field with soldiers for a 30 or 45 days, and then award Small Business Innovation Research contracts for another six months of development.

From there, they could use FUZE’s Tech Maturation Program to buy five, then 20, then maybe 100 pieces of equipment, while working on a Program Objective Memorandum that would lay out a five-year acquisition plan.

“All of that, in maybe a year now, we have moved the needle faster than we ever could, because we would have still been studying the problem” under the Army’s old acquisition model, Daiyaan said.

He couldn’t put a number on the number of different programs PIT is trying to get off the ground, he said, but there are hundreds.

“We're 90 days in—check back with us this time next year,” he said. “We probably should talk about what we transitioned into programs.”

]]>

Move over, Best Ranger; the Army’s looking for the best drone pilots

Meghann Myers — Thu, 19 Feb 2026 14:34:59 -0500

You’ve probably heard of Best Ranger or Best Sapper: Army competitions that test the skills of teams of infantrymen and combat engineers. This year, the service added Best Drone Warfighter.

The inaugural battle kicked off Tuesday at the University of Alabama in Huntsville, bringing teams from across the active, Reserve, and National Guard components of the Army to test their skills and possibly win a slot on the service’s drone competition team.

“At the end of the day, it's not about receiving trophies or awards—it is about ‘what lessons can we take from this to find out who the best operator is and how they became the best operator? What skills and resources and training allowed them to become the best operator?’” Col. Nicholas Ryan, who leads the unmanned aerial systems team for the Aviation Transformation Integration Directorate at Fort Rucker, Alabama, told reporters. “And who's doing some amazing innovation out there across the Army…that we can then take and scale across the entire Army?”

The service is moving away from its previous drone operator model, which trained soldiers in its aviation branch to operate specific platforms. Instead, it’s likely that soldiers with additional training in operating UAS will be integrated into infantry, armor and other frontline units, where new doctrine will have them working alongside machine gunners, Abrams tanks, and howitzers.

“As we proliferate drones, and we're seeing where they best fit into the formation, what we're going through right now is deciding who are the right people to operate these, and what level of training do they need?” Ryan said. “And this competition really helps pull that out. For this competition, we didn't specify what type of soldier—what branch, [military occupational specialty] came here to do this—it was just: ‘Send your best UAS operators’.”

The three-day meet included two different lanes, plus a separate innovation competition where soldiers could submit white papers and custom drone builds, or demonstrate their piloting skills.

The first lane is a race through an obstacle course flying a first-person viewer drone. The second is a hunter-killer scenario, where soldiers camouflaged themselves with paint, dragged a weighted sled and did an overhead water-can press (events similar to the service’s physical-fitness test), then had a half-hour to identify and fire at five high-value targets.

“The first drone is the hunter drone, their reconnaissance drone, and it's looking at an array of targets—about a company-size element of targets—and trying to decide which one out of those are the most important targets. And then the other drone operator is carrying the killer drones, the smaller one-way lethal drones, but they're not kinetically lethal in this case. And then they have to use those to hit those targets.”

Ryan said that while soldiers have been able to execute the movements and operate their drones properly, there have been communication breakdowns as they worked to get into position, identify targets, and fire on them.

“That's an example of something we didn't anticipate, but it's absolutely standing out as that is something we as an Army need to do better on,” he said. “If we're going to proliferate these drones and want them to be more effective and lethal, we just need to improve on how our soldiers talk to each other to communicate when they're using them.”

Units were invited to bring their own small drones to the competition, with no strict rules about the brand, type or capabilities. That also meant they decided which and how much equipment to carry, something the Army is looking to standardize.

“When we're sending soldiers out to carry this equipment as part of a squad or a platoon, and they're carrying it in their rucksack, what is too much?” Ryan said. “How many batteries? How many drones? What types of controllers?”

Can they carry 20 killer drones, or does it make more sense to pack five?

“So kind of developing a standard packing list for a drone operator is one thing out of this competition that we haven't defined or said yet, but we're definitely seeing a range of solutions from soldiers,” he said.

For next year’s competition, officials want to add more realistic scenarios, including the jamming threat that Ukrainian troops are seeing so often.

“We already talked about flying in a congested environment with electronic warfare and building those into the lane,” Ryan said. “And so that's how we're thinking about this: what should we be pushing as a competition that are the highest-priority things our units should be training on to get really good at for their job in the Army?”

]]>

The Pentagon says it’s getting its AI providers on ‘the same baseline’

Patrick Tucker — Wed, 18 Feb 2026 14:34:02 -0500

WEST PALM BEACH, Fla.—After weeks of back-and-forth with AI company Anthropic, the Pentagon is actively talking with all four major U.S. AI players—Anthropic, OpenAI, Google, and xAI—to ensure the companies and the Defense Department are at "the same baseline" regarding Pentagon expectations, the undersecretary of defense for research and engineering said Tuesday.

“We actually signed contracts with all four of them over the summer without a lot of specificity,” Emil Michael told a group of venture capital investors during an Amazon Web Services event. “Now we want to deploy [them] on our system so other people can build agents and pilots, and deploy it,” he said.

In other words, after months of exercises and experiments, the Pentagon is looking to allow different command elements and business entities to build AI agents that can perform a wider variety of tasks with minimal human oversight.

The discussions between Anthropic and the Pentagon have grown increasingly tense. Sources inside the company told Reuters the Defense Department was pushing to use Anthropic’s AI models for domestic surveillance and autonomous weapons targeting, and Axios reported the Pentagon is “close” to cutting ties with the company over Anthropic’s refusal to give the Pentagon unrestricted access to its models. Some Pentagon officials, speaking anonymously, have even vowed to make Anthropic “pay a price” for its perceived lack of cooperation.

Anthropic, which is heavily backed by AWS, is “having productive conversations, in good faith” with the Pentagon, according to a company spokesperson.

Michael struck a far more conciliatory note than other Pentagon officials who have spoken on the spat, and appeared at the event beside AWS Vice President of Worldwide Public Sector Dave Levy.

However, Michael did not move from the Pentagon’s red line. “We want all four of them,” he said, describing OpenAI, Google, xAI, and Anthropic as America’s “AI champions” with the financial staying power for long-term partnership.

Still, Michael noted there are a wide variety of roles the companies might be able to play, and the Pentagon wants different business and command elements to determine what to do with the models, rather than have the companies tell the military what they can and cannot do.

“We're wanting all four companies to hear the same principle, which is: we have to be able to use any model for all lawful use cases.”

Of the four companies the Pentagon has contracted with, Michael said Anthropic is the only holdout on the issue of their ethical safeguards versus the Pentagon’s.

“Some of these companies have sort of different philosophies about what they want it to be used for it or not, but then selling to the Department of War. We do Department of War-like things.”

The Pentagon’s own safety or ethical safeguards must over-rule company safeguards, he said. He described an “extremely dangerous” hypothetical in which the U.S. military could be using an AI agent that suddenly stopped functioning due to embedded company safeguards. “That’s a risk I cannot take.”

The Pentagon has its own safeguards, a list of ethical principles enacted during the first Trump administration that governs everything from development to testing to deployment of AI systems. While the Pentagon’s newest AI acceleration strategy questions the very meaning of “responsible AI,” Michael said adherence to the AI ethics principles is still very much in place.

“The good news/bad news about a hierarchical department is when there's a set of secretary-validated guidance directive memos that lays out the policies and procedures, people follow them. So that's not an issue.”

]]>

Initial Sentinel ICBM expected by early 2030, Air Force says

Thomas Novelly — Tue, 17 Feb 2026 18:18:06 -0500

The troubled Sentinel intercontinental ballistic missile program will hit a key milestone by year’s end, with hopes to deliver the initial ICBM by the early 2030s, Pentagon and Air Force officials announced Tuesday.

In 2024, the Northrop Grumman program to modernize the land-based arm of the nuclear triad went so far over budget that the Pentagon rescinded a 2020 decision to move the program into its engineering and manufacturing development phase. Program officials said in September 2025 that they hoped to re-enter that phase by mid-2027, but now say they plan to hit that milestone this year.

“Leveraging considerable progress over the last 12-18 months, program officials are executing a transformed acquisition strategy paving the way to complete the restructure and achieve a Milestone B decision by the end of 2026, while delivering an initial capability targeted for the early 2030s,” the service said in a Tuesday news release.

Air Force officials said successful ground tests, solid rocket motor qualifications, and critical design reviews are examples of much-needed progress since the program triggered a Nunn-McCurdy Act review in 2024. Additionally, service officials highlighted the new direct reporting portfolio manager role for critical major weapon systems—which includes F-47, B-21, and Sentinel—as crucial to accelerating the program. Gen. Dale White was confirmed for that position in December.

“The DRPM has the direct authority to make decisions, informed by integrated inputs across the enterprise and in alignment with the mission priorities set by the Secretary of War and the Secretary for the Air Force,” White said in the news release. “That construct allows us to resolve tradeoffs quickly and move with the speed required to deliver credible deterrence—while preserving the discipline this mission demands.”

White’s position, and the creation of his new direct reporting portfolio manager role that reports directly to Deputy Defense Secretary Steve Feinberg, appeared to be at odds with Defense Secretary Pete Hegseth’s acquisitions reforms—including changes to expand decision-making authority at lower levels.

The new role “signifies a major lack of confidence in the Air Force to pull off its main acquisition programs” and allows White and the Pentagon to call the shots on Sentinel, said Todd Harrison, a defense budget expert at the American Enterprise Institute.

“I think it also signifies that there are certain programs that are too big or too important to fail,” he said. “A lot of these processes it doesn't seem like are going to apply to these too-big-to-fail programs. Sentinel is definitely one of those, because we don't have an alternative. We don’t have a fallback.”

Other Sentinel program developments:

Construction has already started on the first of three new command and control centers at F.E. Warren Air Force Base in Wyoming, and on test facilities at Vandenberg Space Force Base in California.
Teams plan to break ground on a prototype launch silo at Northrop Grumman’s Promontory, Utah, site, this month, which “will allow engineers to test and refine modern construction techniques, validating the new silo design before work begins in the missile fields,” the news release said.
Prototyping activities this summer at F.E. Warren will validate construction methods being used for utility corridors.

Officials are planning for the first missile pad launch for Sentinel in 2027, the news release said.

“The deliberate progress being made on Sentinel ensures that for decades to come, there will be no doubt in the minds of our adversaries about the credibility and readiness of our nation's nuclear deterrent,” Navy Adm. Richard Correll, head of U.S. Strategic Command, said in the news release. “That is the ultimate deliverable."

]]>

The Army is writing the book on using small drones in a tank formation

Meghann Myers — Tue, 10 Feb 2026 17:51:51 -0500

As the Army’s Transformation-in-Contact brigades test and help develop new technology, they’re also shaping how soldiers will be trained to use it.

At Fort Stewart, Georgia, soldiers in the 3rd Infantry Division are working on a pair of courses to certify soldiers to operate small unmanned aerial systems, part of a servicewide effort to create doctrine around using drones throughout every formation.

“3rd ID, specifically, is developing ways to qualify their operators on the different systems, and we are sending feedback back through the proper channels to big Army, to work on developing an Army-wide qualification course,” Capt. William Langley, who leads the UAS and electronic warfare element in the 2nd Armored Brigade’s 6th Squadron, 8th Cavalry Regiment, told reporters Tuesday.

Both the 1st and 2nd Armored Brigade Combat Teams are participating in TiC, with a special focus on integrating drones into tank units, said Capt. Brenden Shutt, the division’s innovation officer.

“The legacy UAS systems were focused on dedicated 15-series UAS operators, whereas now, we're leaning more toward training standard infantry and armor soldiers to be the UAS operators,” Shutt said.

So far the division has been focusing on UAS classified in groups 1 and 2—that is, under 55 pounds. They’ve fielded more than 150 of them, Langley said, including the Anduril Ghost-X, Performance Drone Works C100, the Neros Archer, the Teal 2, and the AeroVironment Switchblade 600.

“We mainly use these in a hunter-killer format, so we use the recon assets to find the targets, and then Archers…to strike the target,” Langley said.

To qualify, soldiers start with about 40 hours of simulation training before ever touching the UAS itself.

“After that, we work with smaller [first-person view drones], and they have to hit some obstacle courses and hit some gates before they go to a larger FPV, where they then actually learn to fly them, and then put certain skills into practice and learn to integrate them into conventional armor/infantry tactics,” Langley said.

And rather than picking out a handful of soldiers to be the designated UAS operators, 3rd ID wants everyone to be familiar with flying a drone, in the way they are all qualified on their rifles.

“Previously, it was very much a select group of people who were tasked with leveraging this technology to deliver effect,” Shutt said. “Now, every single soldier, from a maintainer to an infantryman to a signal soldier who fixes the radios—all of them kind of have a baseline understanding of how drones work, and could, with the sim time, with a little bit of minimal field training, be able to deliver effects during an operational or training environment.”

]]>

Pentagon leaders should have more control over services’ tech budgets, GAO suggests

Patrick Tucker — Thu, 05 Feb 2026 19:14:02 -0500

The Pentagon could further accelerate its technology purchasing if the services’ emerging-tech budget requests flowed through the office of the defense undersecretary for research, the Government Accountability Office says in a new report.

The report urges lawmakers to give "budget certification authority" for the services’ research and engineering spending to the Office of the Under Secretary of Defense for Research and Engineering.

“This would require the secretary of each military department and the head of each defense agency to transmit their department’s or agency’s proposed budget for research, development, test, and evaluation activities,” the report says. The R&E office would then “review each proposed budget and determine whether it is adequate.”

Unsurprisingly, the proposal was not well received by the services.

“The Departments of the Army, Air Force, and Navy disagreed,” arguing that the change would lead to “delays, restricted autonomy, and increased workload,” the report said.

But GAO says the current setup limits the Pentagon tech chief’s ability to ensure that service purchases fit with broader plans for the joint force—a “key role” the office was intended to play.

Consolidation, happening

The Pentagon has already pushed through a variety of measures to speed up and coordinate technology efforts. A March memo prioritizes the purchase of existing “dual-use” technology, particularly software, over custom, service-built solutions. The Department also is pushing acquisition authority down to the tactical level, allowing colonels and Navy captains to buy equipment in small batches using other transaction authorities.

These moves have met with approval from long-time Pentagon watchers like Paul Scharre, author of Four Battlegrounds and vice president of the Center for a New American Security.

“This leadership team is very invested in shaking things up, moving faster, and clearing out some of the red tape,” Scharre told Defense One in December. “A big piece of this has to happen on Capitol Hill as well. Congress has to be supportive...that has to do with things like getting less control out of the appropriators and giving the Department of Defense more flexibility to spend money very fast and be flexible in how they move money around.”

The NDAA fight

The current draft of the 2026 National Defense Authorization Act, passed by the Senate in October, includes a provision that echoes the GAO’s recommendation but stops short of full certification power.

Instead, the Senate-passed bill would establish Portfolio Acquisition Executives, or PAEs, to replace traditional Program Executive Officers. These PAEs would have the ability to change requirements on their own and would grant the Pentagon’s research office more direct authority over their activities.

However, the House version of the bill is more cautious. It stipulates that these acquisition executives “shall continue to report through their respective functional commands.” This discrepancy remains a primary point of contention as the House and Senate move to reconcile their versions of the bill. Despite the friction, there is at least a philosophical agreement that the services must buy equipment that aligns with a broader joint-force strategy.

The GAO report also included a pointed note for the Pentagon’s tech leaders: the office “has yet to ensure that Critical Technology Area roadmaps consistently provide sufficient information for military departments to invest in technologies for the joint fight.”

In other words: if the Pentagon is expected to follow a comprehensive tech strategy, the tech chief needs to finish writing it down.

]]>

New Pentagon science-and-innovation board arrives as administration cuts research funding

Patrick Tucker — Tue, 03 Feb 2026 05:03:30 -0500

The Pentagon's new Science, Technology, and Innovation Board—a merger of the decade-old Defense Innovation Board and the 70-year-old Defense Science Board—is meant to “streamline” the department's approach to the hardest technological and scientific national-security challenges. But it comes on the heels of Trump-administration cuts that could hinder those efforts.

Streamlining is a persistent target for the Pentagon. But it’s one that it has had trouble achieving in previous years, according to GAO reports, lawmakers, and military leaders across administrations. It is one reason why the so-called “valley of death,” as in the chasm between a cutting-edge research program and an actual weapon getting into the hands of soldiers, remains a common complaint—and one of the key reasons the Defense Innovation Board was created in the first place.

The DIB, established in 2016, was a civilian body whose members had included tech and finance leaders such as Eric Schmidt, Michael Bloomberg, and Neil deGrasse Tyson. It aimed to bring thought leadership from top business leaders into the Pentagon. The board produced a wide variety of key recommendations that the Pentagon later adopted, such as moving to large-scale enterprise cloud computing and adopting a long list of ethics principles for the development, testing, deployment, and operation of artificial intelligence across the military.

The Defense Science Board, meanwhile, largely produced reports for Congress and military leadership on specific Defense Department issues, such as how to reform testing and evaluation and bringing more digital engineering into the department.

The new Science, Technology and Innovation Board, or STIB, includes defense science experts in areas such next-generation autonomy, testing, advanced hypersonics, and acquisition. It also includes private-sector experts in fields such as advanced neural networks.

The new board emerges at a time when the military is keen to integrate artificial intelligence into more of what it does, reach new research breakthroughs more rapidly, and quickly produce large numbers of cheap, highly autonomous drones.

One former senior defense official said the board should look at “all aspects of AI, from energy requirements, ethics, and direction of research,” to “how to accelerate the fusing of the massive amounts of all-domain sensory data the department has to train multimodal AI for offensive and defensive operations.”

However, they pointed out that the board is uniformly white and largely male. “It misses the mark as far as representation goes, thereby handicapping its credibility with the American public,” they said.

Another question is whether the new board will conduct public meetings, as the Defense Innovation Board did, or meet in private, like the Defense Science Board. The STIB announcement makes no mention either way, and a Pentagon spokesperson did not respond to a request for comment Monday.

Shrinking the defense science, research, oversight footprint

The new board is the latest in a series of Pentagon moves to merge offices or activities launched over the last decade and accelerate the adoption of AI, particularly dual-use AI from companies that also sell to the public.

On January 12, the Pentagon announced an effort to accelerate the use of large foundation models such as Google’s Gemini. They simultaneously announced an “overhaul” to more closely align offices like the Chief Digital and Artificial Intelligence Office, the Strategic Capabilities Office, and the Defense Innovation Unit under the under secretary of defense for research and engineering.

These moves align with what current and former military officials, government watchdogs, and lawmakers have long been urging Pentagon leaders to do: reform the way the department buys technology to be more like DIU. The reorganization of these offices and activities was taken by many observers across the political spectrum as a sign that Pentagon leadership had finally begun to do just that, and was on its way to busting down bureaucratic obstacles to buying ready-to-use commercial technology, especially software.

But the DSB-DIB merger comes as the Trump administration reduces funding for basic sciences. According to the most recent version of the 2026 National Defense Authorization Act, the department would cut support for basic research at U.S. universities by nearly 5%.

At the same time, the Pentagon is de-emphasizing, if not abandoning, the ethical use of AI. The AI-acceleration strategy released in January does not even mention the AI ethics principles the Defense Innovation Board proposed, and the Pentagon adopted, back in 2020. Instead, it advises the Defense Department to “incorporate standard ‘any lawful use’ language into any [DOD] contract through which AI services are procured within 180 days.”

The department’s ability to determine lawful use is also shrinking, at least at the highest levels, following the replacement of several top JAGs last February and the sidelining of other JAG officers when considering controversial moves, such as firing at unarmed boats or deploying the National Guard for immigration enforcement. The department has also significantly reduced its Office of Inspector General, which provides key oversight on issues like safety and policy effectiveness.

All of that is causing confusion and disagreement with some of the very AI companies the Pentagon is courting. Researchers at Anthropic, for instance, are concerned the Defense Department might ask them to change or modify AI tools against company guidelines in order to fit whatever definition of “lawful use” the Pentagon is working with at the moment, according to reporting from Reuters.

One current official said the public reporting on the disagreement between the department and Anthropic overstated the issue somewhat, and that the new policy simply says the Pentagon “should be the entity dictating lawful use and safeguards and not having companies specifying how products will be used within cases of lawful use. The conversation of who will write those safeguards is separate.”

A former senior defense official who worked on areas like deployed artificial intelligence agreed that some of the concerns about the Pentagon’s new approach to AI safety were a tad panicky.

However, they said, a larger issue is the testing and evaluation standards the Pentagon and the services use to field AI software. “As long as those standards remain high, the policy stuff surrounding them is window dressing... because terrible, unvetted software won't get scaled and fielded.”

But the Pentagon has also been reducing the amount of money and staff dedicated to ensuring that testing and evaluation remains rigorous, essentially shrinking the office that oversees service testing by half last May. But the former official said that because the services themselves do the initial testing, and are far more committed to fielding safe AI that won’t harm service members than a Pentagon office would be, that reduction in and of itself doesn’t present a problem.

Still, while any particular merger or reduction may not be a cause for alarm by itself, the totality of the shrinking could be. Might these cuts and mergers affect core areas the Pentagon depends on to set standards for things like deploying AI? That’s worth keeping track of, said the second former official.

“If those standards decline, you have problems. The number of people you have doing AI safety, in particular, is potentially related to that, but not necessarily,” they said. The Pentagon leadership was right to cut the “waste and duplication in the [testing and evaluation] process,” they said. “And I also think they are probably implementing it in the dumbest way possible.”

]]>