The U.S. Department of Defense is pushing an "AI-first" strategy to modernize military operations, but a critical gap exists between the Pentagon's enthusiasm for artificial intelligence and soldiers' willingness to rely on it in the field. Without overhauling how the military tests and evaluates AI systems, experts warn that unpredictable AI behavior could erode troop confidence and create dangerous blind spots in combat situations.

What's Driving the Pentagon's AI Push?

The Pentagon has invested heavily in its Artificial Intelligence Strategy this year, explicitly aiming to reward AI-driven "re-conceptions of legacy approaches" across military operations. AI already plays a significant role in current military activities. Palantir's Maven Smart System, a Florida-based defense tech firm's platform that uses multiple forms of AI including frontier models, helped identify more than 13,000 targets in just over the first month of military strikes against Iran. The military sees AI as essential to maintaining technological advantage and accelerating decision-making in an increasingly complex threat environment.

The Army is investing heavily in the next level of AI integration and increasingly having military contractors responsible for AI development work directly with troops. This new model for solving military problems is speeding up the process, but it doesn't guarantee that soldiers will actually trust or use these systems effectively.

Why Are Soldiers Hesitant to Trust AI Systems?

The core problem isn't that AI doesn't work; it's that AI systems often behave inconsistently. Jake Steckler, a research scholar at GovAI, a nonprofit think tank focused on AI research, has studied how military personnel interact with AI tools. He found that AI-enabled systems frequently produce different responses or choices even when only small changes occur in their environment or the data used to train them.

"Soldiers need to see these things working consistently for them to trust them. It'll take time for anyone to be able to build that trust and I think there's a lot of technical problems that still need to be solved before that trust is even possible to build," said Jake Steckler.

Jake Steckler, Research Scholar at GovAI

This unpredictability creates a credibility problem. When a soldier encounters an AI system that behaves differently under similar conditions, they naturally become skeptical about relying on it in high-stakes situations. The risks are compounded because these technologies are so new that unexpected scenarios could emerge in the field, potentially leading troops to abandon AI tools altogether.

How Is the Pentagon's Organizational Structure Contributing to the Problem?

The Pentagon has consolidated its AI purchasing process to reduce bureaucracy and speed up deployment of modern tools to service members. However, this streamlining has created organizational blind spots. Responsibilities that previously existed across multiple departments are now unclear in terms of ownership and accountability.

"Getting rid of an old process is good. But you have to replace it with a new, better process; otherwise, you're just creating confusion and introducing new risk," explained Jake Steckler.

Jake Steckler, Research Scholar at GovAI

This consolidation loads a single office with specialized responsibilities that were once distributed throughout the military structure. Without clear ownership of testing, evaluation, and oversight functions, the Pentagon risks deploying AI systems that haven't been adequately vetted for real-world military conditions.

Steps to Building Soldier Confidence in Military AI Systems

• Modernize Testing Standards: The Pentagon must develop comprehensive testing and evaluation protocols specifically designed for AI systems, moving beyond legacy approaches that were built for traditional military equipment and don't account for AI's unique failure modes and inconsistencies.
• Ensure Consistent Performance: Military AI systems need to demonstrate reliable behavior across varying environmental conditions and data inputs before deployment, so soldiers can predict how the system will respond in different scenarios.
• Clarify Organizational Accountability: The Department of Defense should redistribute oversight responsibilities across multiple specialized teams rather than concentrating them in a single office, ensuring that testing, evaluation, and deployment decisions receive proper scrutiny.
• Involve Troops in Development: Military contractors should continue working directly with soldiers during AI development and testing phases, incorporating frontline feedback to identify inconsistencies and usability issues before systems reach operational units.

What Can the U.S. Learn From Ukraine's AI Battlefield Experience?

Ukraine has become the world's leading testing ground for AI-enabled military systems, including one-way drones and autonomous munitions. The country's approach offers valuable lessons, though they come with important caveats. Ukraine accelerated its technology approval timeline from roughly two years to as little as two weeks after Russia's invasion, prioritizing speed over thoroughness.

Steckler, who informed his research through discussions with troops in Ukraine and his own experience as an Army helicopter pilot, noted that Ukraine's willingness to "prioritize speed over thoroughness" works because the country faces an existential threat. When the choice is between fielding an imperfect system or surrendering territory, soldiers are motivated to accept higher risk.

However, this approach has limits. More complex AI systems haven't been adopted in Ukraine because, as Steckler observed, "humans still outperform the available systems." Additionally, Ukraine's success depends on mobilized public support, with citizens working nights in factories or assembling circuit boards at home. The U.S. military operates in a different context without the same existential pressure, meaning American troops may not accept the same level of risk or sacrifice.

Are AI Companies and the Pentagon Aligned on Military Use?

A significant tension has emerged between some leading AI firms and the Pentagon over how military AI systems should be deployed. Anthropic, a California-based AI company, refused the Pentagon's demand for unrestricted use of its most powerful AI large language models. In response, the Trump administration banned Anthropic's use and labeled the firm a national security supply-chain risk, a designation that could severely damage the company's ability to do business with the federal government.

This dispute highlights a broader challenge: the Pentagon wants maximum flexibility in how it uses AI systems, while some AI developers want restrictions on military applications. This legal and policy conflict is still moving through the U.S. court system and could shape how the military accesses advanced AI capabilities in the future.

The Pentagon's AI-first strategy is ambitious and necessary for maintaining military competitiveness. But without solving the trust problem, the military risks deploying powerful tools that soldiers won't actually use when it matters most. Building that trust requires not just faster technology, but better testing, clearer accountability, and honest conversations with troops about what AI can and cannot reliably do in combat.