Artificial intelligence isn't just coming onto the battlefield. It's a full-fledged structural change. There are bigger plans, from sensors turning into decision-making nodes to communication links turning into adaptable networks. Moreover, autonomy, in addition to increasing power, also brings responsibility. The United States currently faces a dual challenge. How to leverage the transformative potential of AI while managing both tactical and strategic risks.
Reprioritizing the Department of Defense: Increase focus and reduce dispersion
In August 2025, the Department of Defense announced plans to reduce the list of “critical technologies” it monitors, narrowing the scope rather than expanding it. To avoid dilution, we get lazy about calling any emerging technology “important” and then spread it too thin. The new stance indicates that AI, hypersonics, and directed energy are likely to be prioritized among the three elements.
Military programs, contracts, and research and development will be subject to review. The impact is significant. Projects that go beyond the traditional core may be difficult to obtain funding or exist only as part of a larger project.
AI at the tactical edge: communications, networks, and autonomy
New research points to a fundamental change. Defense networks and tactical communications are becoming more than just links, they are becoming AI-enabled systems. A study reported in 2025 will help engineers showcase the use of AI in tactical communications with autonomous routing. Essentially, this network is both a military nervous system and a semi-autonomous organ that adapts to pressure.
However, this comes with risks. Adversarial AI attacks disrupt the stability of the entire network. Autonomous measures and miscalculations in the decision-making loop may occur. According to one study, ML (machine learning) is already enabling the replacement of human combatants with AWS (autonomous weapons systems). The use of such ML-controlled weapons increases the likelihood of low-intensity conflict, which can escalate even without full human control.
Therefore, the United States should establish rigorous AI-ready metrics and ensure that systems are transparent, testable, predictable, and (where appropriate) human-involved.
Manned-unmanned teaming: increasing force and managing risk
MUM-T is a promising field that bridges humans and machines. Instead of full autonomy, the concept proposes a human platform controlled by a drone “winger” under cooperative control. The Anduril Fury and General Atomics Gambit prototypes have been developed in a race to support the U.S. Air Force's efforts to develop manned fighter jets.
Although MUM-T still relies on human judgment, it spreads risk and expands the scope of application. The line between human override and autonomous response is narrow and many, so transitions in command, control, and decision-making must be tightly defined.
Supply chain is the weak link: semiconductors, sensors, trust
AI systems rely on semiconductors, sensors, and software stacks. Without a resilient and reliable supply chain, the integrity of the entire system becomes questionable. The United States responded to this threat with the passage of CHIPS and the Science Act. This refers to $52 billion in subsidies and other incentives aimed at bringing chip manufacturing back home to reduce dependence on geopolitical rivals.
But this industrial gamble is not guaranteed. Apart from the chip, there are also concerns about lidar sensors. The report suggests that China's LIDAR systems have hacking and backdoor vulnerabilities that could pose a threat. The quality of the system is also subject to speculation.
Lessons learned from this study show that the hardware base influences the reliability of perfect AI algorithms.
Escalation risk and the AI arms race
Great powers tend to become too competitive. In military systems, lower barriers to use, such as faster decision-making loops and automated responses, can undermine crisis stability. In 2024, the United States took the lead in publishing the Political Declaration on the Responsible Military Use of AI and Autonomy. As of 2024, 51 countries have signed this declaration. However, compliance is voluntary and enforcement is weak.
Armed systems that use AI without guardrails can lead to unintended escalation. There is very little difference between winning and losing.
Policy imperative: What’s next for U.S. strategy?
Adopt an AI-enabled framework: Traditional Technology Readiness Levels (TRLs) do not take into account the uncertainties and risks of AI. Better frameworks should account for model drift, adversarial robustness, interpretability, failsafe behavior, etc. Recent research calls for just that.
Networked testbeds and fighter pilots: Accelerating the use of real-world test settings to test AI systems in competitive environments. Operational feedback must feed into the design loop.
Strengthen your supply chain and source transparency: Monitor where all critical subsystems come from, ensure trusted builds, and maintain multiple back roads. Avoid or heavily audit entities that rely on a single supplier in a hostile country.
Rules of engagement and fail-safe protocols: Human overrides, kill switches, and strict escalation peering must be non-negotiable in conflict zones. In high-risk areas, AI-powered systems should not operate completely autonomously without multiple layers of oversight.
International norms and confidence-building: Verification, red lines, and shared protocols need to be strengthened at the treaty level (not just symbolic declarations), especially with peer companies like China and Russia.
conclusion
America has a choice. Either AI will give the country an advantage in the future, or it will create long-term liability through overreach and inappropriate participation. Victory does not necessarily lie in maximizing the autonomy of defense systems, but rather in how to integrate autonomy into systems in a disciplined, strategic, reliable and safe manner. This points to the need for smarter acquisitions, more secure architectures, stronger supply chains, and better risk management.
If the Pentagon can combine these elements—and in other words, innovate recklessly—it may still maintain its historic advantage.
