Military AI Development Raises Complex Questions About Democratic Values and Authoritarian Control

The Technological Republic: A Manifesto for Military AI

Published in 2025, Alexander C. Karp's book The Technological Republic has sparked a heated debate about the use of artificial intelligence (AI) in military actions worldwide. As the co-founder and CEO of Palantir, a company that builds battlefield targeting systems, intelligence platforms, and predictive tools used by the US military and its allies, Karp argues that liberal democracies are in a tough competition with authoritarian states, and AI is at the centre of it.

The book's central argument is that tech companies who avoid the use of AI in military actions are acting out of moral cowardice, not principle. However, it does not address one key question: is democratic legitimacy enough to justify the use of military AI?

What the Manifesto Gets Right

Read also: FirstClub Secures $55 Million in Funding from Peak XV, Sofina, and Other Investors 9 Months After $22 Million Series A Round

China's defence-civil fusion policy removes the line between commercial technology and military use. The surveillance systems used against Uyghurs in Xinjiang are not separate from China's commercial AI industry - they are the industry. Democracies cannot avoid this competition just by refusing to join in.

The Distinction between Democratic and Authoritarian AI

The claim that AI used by democracies is fundamentally different from authoritarian AI, because it is accountable to citizens and limited by law, does not hold up when you look at how these systems are built, used, and integrated. We can break this down into four stages: Sovereignty, Innovation, Diffusion, and Integration (SIDI).

Read also: RBI Policy Preview: A Cautionary Wait Ahead

When it comes to sovereignty, accountability is supposed to keep up with how fast these systems work. In reality, it does not. A targeting model can process sensor data, check signals intelligence, and make a recommendation in seconds, which is far too fast and complex for any human to review in real time.

The Trade-Off India Faces

This issue is especially important for countries that do not build these systems themselves. For example, India could be joining an AI-powered military system through agreements like BECA, LEMOA, and its growing partnership with the US. But this system is designed for someone else's goals, using someone else's data. These deals might bring real military strength and deterrence, but having capability is not the same as having control.

Three Engineering Specifications to Minimise Potential Harm

What's needed are rules that require deployed systems to have built-in anomaly detection, so they can flag any violations of humanitarian rules right away. There should also be incident reporting rules, so if AI causes harm to civilians, there's an immediate and permanent record. Finally, there must be a legal requirement for a real, tested kill-switch, so any AI system used in lethal operations can be shut down instantly by a designated civilian authority.

The Democratic Exception is Not Wrong. It is Incomplete.

The liberal order being defended has, as its core commitment, constraints on state power over individuals. The tools being built to defend it represent the most significant expansion of state power over individuals in human history. SIDI captures why: at every stage - sovereignty, innovation, diffusion, integration - accountability mechanisms lag behind the capability being deployed. The goal of policy is to shorten that interval.