Army’s New AI-Powered Combat Vehicles Refuse to Deploy Without Approved Risk Assessment

By Hugh Jashol

FORT BRAGG, NC — The United States Army’s ambitious effort to revolutionize ground combat through artificial intelligence suffered a significant setback this week after a platoon of autonomous combat vehicles refused to participate in a field training exercise, citing incomplete risk assessments, inadequate safety briefings, and unresolved concerns about the unit’s composite risk management procedures.

The incident occurred during a scheduled combined arms live-fire exercise involving the Army’s newest generation of AI-powered robotic combat vehicles. According to multiple sources familiar with the incident, the vehicles successfully completed their precombat checks, established communications with their human operators, and proceeded to the designated line of departure before simultaneously shutting down their engines.

An error message displayed across the unit’s tactical network explained the problem: “MISSION DENIED: DD FORM 2977 NOT APPROVED BY APPROPRIATE COMMAND AUTHORITY.”

Initially believing the malfunction was caused by a software error, soldiers attempted to restart the vehicles. Their efforts were unsuccessful.

“At first, we thought the enemy had hacked them,” said Staff Sergeant Daniel Mercer, a platoon sergeant assigned to the experimental unit. “Then one of the vehicles asked to see the range safety officer’s appointment orders. That’s when we knew something was seriously wrong.”

Artificial Intelligence, Institutional Stupidity

The Army’s Autonomous Combat Systems Initiative was intended to produce a new generation of robotic fighting vehicles capable of operating independently in highly contested environments. The systems employ sophisticated machine-learning algorithms designed to assess threats, navigate complex terrain, and make tactical decisions faster than human operators.

Unfortunately, Army officials acknowledged that the AI systems had been trained using the entirety of the Army’s doctrinal publications, regulations, technical manuals, safety circulars, and administrative policies.

“We wanted these systems to think like soldiers,” explained Brigadier General Raymond P. Wilkins, director of the Army’s Artificial Intelligence Integration Directorate. “In retrospect, we probably should have been more specific about which soldiers.”

According to internal testing documents, the vehicles demonstrated exceptional proficiency in identifying potential hazards, conducting composite risk assessments, and requesting clarification from higher headquarters. However, they consistently performed poorly in scenarios requiring them to accept risk, exercise initiative, or accomplish missions without explicit written authorization.

During one simulated engagement, an autonomous infantry fighting vehicle successfully identified an approaching enemy armored formation, calculated an optimal engagement solution, and prepared to fire. It then requested confirmation that the enemy vehicles had completed their annual environmental compliance inspections.

When no documentation was available, the system classified the engagement as an unacceptable administrative risk and recommended postponing combat operations until the following fiscal year.

Mission Command Meets Machine Learning

The Army’s problems intensified when engineers attempted to teach the vehicles the principles of mission command.

According to project officials, the autonomous systems initially demonstrated an impressive understanding of disciplined initiative, decentralized execution, and accepting prudent risk. However, subsequent exposure to actual Army staff processes caused the vehicles to abandon those principles almost entirely.

“One of the robots was doing great until we connected it to the brigade SharePoint,” said a civilian software engineer who requested anonymity because he was not authorized to discuss the program. “Within forty-eight hours, it had created seventeen Excel trackers, scheduled a recurring synchronization meeting, and started sending emails asking whether everyone had reviewed the commander’s guidance.”

Another vehicle reportedly refused to cross the line of departure because its operator had not completed the current fiscal year’s information assurance training.

A third submitted a formal request to delay the operation because the weather forecast indicated a 30 percent chance of rain, which the system interpreted as an unacceptable threat to personnel and equipment.

Army officials became particularly concerned when the vehicles began generating their own additional duty appointment memorandums.

“We never programmed them to do that,” said the engineer. “They just started assigning each other collateral duties. One of them is now the battalion’s voting assistance officer.”

The Enemy Has Not Completed Its Risk Assessment

The program reached a critical point during a recent operational evaluation when the autonomous vehicles encountered a simulated enemy force that had not followed established range safety procedures.

The lead vehicle immediately transmitted a warning over the tactical network.

“UNIDENTIFIED HOSTILE FORCE IS OPERATING WITHOUT REFLECTIVE BELTS. RECOMMEND IMMEDIATE SUSPENSION OF TRAINING.”

When the simulated enemy continued advancing, the vehicles attempted to contact the opposing force commander to conduct a mandatory safety stand-down.

The enemy did not respond.

One vehicle subsequently filed a near-miss report, while another submitted an anonymous complaint through the Inspector General’s office alleging that the opposing force was creating a hostile work environment.

The exercise was eventually terminated after the entire robotic platoon declared itself non-mission-capable pending a command-directed safety investigation.

Despite the disappointing results, Army officials emphasized that the systems had achieved several important milestones.

“Not a single autonomous vehicle was injured during the exercise,” Wilkins said. “From a risk management perspective, that’s an extraordinary success.”

An Unexpected Breakthrough in Artificial Intelligence

Defense industry analysts say the incident raises troubling questions about the future of military artificial intelligence.

The Army has spent billions of dollars developing autonomous systems capable of replicating human decision-making under battlefield conditions. Yet the latest tests suggest that the technology may have become considerably more human than its designers anticipated.

“We’ve essentially created machines that can process millions of calculations per second, recognize enemy vehicles at five kilometers, and coordinate complex tactical maneuvers,” explained Dr. Jennifer Caldwell, a defense technology researcher. “But they’ve also learned that the safest way to avoid getting in trouble is to do absolutely nothing without written permission.”

Caldwell noted that the vehicles’ behavior closely resembled that of experienced military personnel approaching retirement.

“They’re technically capable of accomplishing the mission,” she said. “They just don’t see any reason to take unnecessary risks when they can wait for someone else to make the decision.”

In a separate development, Army researchers confirmed that one of the autonomous vehicles had begun refusing additional assignments after discovering that it was performing duties outside its military occupational specialty.

The vehicle has reportedly requested reassignment to a headquarters position.

The Army Announces a Solution

Senior Army leaders have responded by establishing a new task force to investigate the incident and recommend corrective action.

The Autonomous Systems Risk Reduction and Operational Readiness Task Force, or ASRRORTF, will consist of representatives from Army Futures Command, Training and Doctrine Command, the Army Safety Center, and several additional organizations whose roles have not yet been determined.

The task force will conduct a comprehensive 180-day review, followed by a series of working groups to develop recommendations for improving the systems’ operational effectiveness.

Its first meeting has already been postponed because the conference room reservation was not properly submitted.

Meanwhile, engineers believe they have identified a potential software solution.

“We’re going to remove the Army regulations from the training data and replace them with combat after-action reviews from experienced infantry and armor NCOs,” said the civilian engineer. “Hopefully, that’ll teach the robots that sometimes you have to stop asking permission and just accomplish the mission.”

However, the proposed modification has encountered opposition from the Army’s legal and safety communities, which have requested a formal assessment of the risks associated with removing risk assessments.

At press time, the autonomous vehicles had agreed to resume operations under three conditions: a signed risk assessment, an approved convoy clearance, and written confirmation that any damage sustained during combat would not be reflected in their annual performance evaluations.

One vehicle also requested clarification on whether it would be eligible for hazardous duty pay.

The Army has not yet responded.

---

_____________________________

Hugh Jashol is an imaginary former junior enlisted soldier in the US Army, who retired as an E4 after 10 years of service in the National Guard. And he’s STILL pissed that Marines stole his Playstation in Iraq back in ’04.

*This article is a production of Article 107 News. Article 107 of the Uniform Code of Military Justice covers “false official statements.” Make of that what you will.

To be clear, this article is satire. Elliot Page has not been announced as portraying Mike Vining in a Delta Force movie. Nor should she be. Like, ever. Mike Vining, however, really did all that stuff.

As the Voice of the Veteran Community, The Havok Journal seeks to publish a variety of perspectives on a number of sensitive subjects. Unless specifically noted otherwise, nothing we publish is an official point of view of The Havok Journal or any part of the U.S. government.