AI military dominance should not be a partisan political issue

The shift to AI warfighting

Artificial intelligence is no longer a theoretical concept in modern warfare; it is actively shaping the battlefield. In the conflict in Ukraine, AI-assisted drone technology has become a primary tool for military operations. These systems allow for high-precision strikes and the ability to hold ground against conventional forces.

Recent data shows the scale of this shift. During March, drones were responsible for ninety-six percent of the more than 35,000 Russian casualties recorded by Kyiv. Furthermore, a Russian position was recently taken by ground robots and drones working together without any infantry involvement.

The People's Liberation Army in China is also prioritizing these capabilities. They have conducted exercises involving more than 200 AI-enabled drones that can coordinate among themselves. This suggests that AI is becoming a decisive factor in how modern wars will be won or lost.

Why voters lack the context

Despite the high stakes of military AI, the technology is often viewed by the public through the lens of everyday inconveniences. Many Americans express anxiety regarding data center construction, job automation, and rising electricity bills. These visible issues often overshadow the classified and technical nature of military AI applications.

As reported by The London School of Economics and Political Science, the complexity of AI policy makes it an unsuitable subject for general electoral debate. Decisions regarding compute requirements to maintain an advantage over China are not matters of voter preference. They depend on technical thresholds and the secret capabilities of adversarial nations.

The author argues that a campaign focused on AI would likely only reinforce public anxiety without addressing core security needs. Because the relevant facts are often classified, voters in local districts cannot know what capabilities are being fielded by global rivals. This makes it difficult for the electorate to make informed judgments on complex export controls or grid allocations.

The link between civilian and military AI

A significant challenge for regulators is that civilian and military AI development are inextricably linked. The same engineers, silicon, and models used to generate social media content are also used to identify targets in drone feeds. Therefore, imposing heavy regulations on private model developers would inevitably slow down the tools the military now relies on.

Private firms have already demonstrated the speed of this development cycle. In Ukraine, private manufacturers now produce around ninety percent of Ukraine’s FPV drones, replacing designs every week to stay ahead of the enemy. In the United States, companies like Palantir and Anduril move quickly because they are not waiting for government approval processes.

If the government seeks to throttle AI development for safety or domestic reasons, it risks harming the very companies providing a lifeline to defense. There is no mechanism to slow down one side of the industry while leaving the other untouched. Restricting power to data centers, for example, would limit the compute available for training both commercial and defense systems.

The risk of unilateral restraint

The author warns that the United States should not assume that its own regulations will influence the behavior of Russia or China. China's military doctrine already treats AI as central to its warfare strategy. There is no evidence to suggest that Beijing would slow its programs because the U.S. chooses to impose domestic restrictions.

Unlike nuclear weapons, software is not scarce and cannot be easily verified by international treaties. The author notes that the PLA has already been found seeking Nvidia chips despite American export restrictions. Because verification is nearly impossible, unilateral restraint may only hand a military advantage to states with less interest in transparency.

The argument is that pacing the frontier in California does nothing to reduce global risks. Instead, it may simply decide who arrives first at a dangerous technological threshold. The author suggests that the only way to manage these risks is to maintain a technological lead over adversarial powers.

Focusing on conduct over development

The solution proposed is to distinguish between policing AI conduct and slowing down its development. The United States can still implement aggressive actions against specific harms, such as fraud, deepfakes, and risks to children. These are issues of conduct that can be legislated for without fundamentally constraining the underlying technology.

The author argues that the government must avoid mindlessly slowing the industry to win votes, whether through stopping data centers or creating rules that hinder the type of competition that helped Ukraine. The goal should be to regulate how AI is used rather than limiting its growth.

The author concludes that the United States must maintain its military capacity to enforce a rules-based order. Because that capacity now rests on AI, letting AI policy be set by electoral politics could lead to a significant disadvantage in critical regions like the Taiwan Strait.