Kingsmith, A. T. (2026). From rule of law to rule of algorithm: Generative Artificial Intelligence’s threat to democracy. Big Data & Society, 13(2).
What drew me to this research was a fundamental problem: the same technology people use to write emails or plan holidays is now deciding who receives welfare, whose visa application succeeds, and what citizens are told their government’s policies mean for them.
Generative AI systems like Microsoft’s Azure OpenAI are already embedded in government services across many countries. These technologies are drafting policies, handling benefit inquiries, and generating citizen communications. My main concern is that this represents a shift in governance procedure and logic from traceable rules and human deliberation toward outputs that nobody fully controls and few can meaningfully scrutinize.
The algorithmic systems of the 2010s (Australia’s Robodebt, the Netherlands’ SyRI) were troubling because they removed human judgment from key decisions. But citizens were still formal legal subjects. There was something to appeal, someone to confront. What changes with generative AI is that the system does not merely classify citizens, it shapes the world they encounter: the explanation of a policy, the framing of a consultation, the letter that nudges someone away from a benefit claim they were entitled to make.
Canada’s Chinook immigration system illustrates the accountability problem. When the tool produced disproportionate rejection rates for applicants from African countries, no one could explain it. The department blamed the tool; the developers said it only assisted human decisions. Applicants had nowhere to go. This is not a peripheral outcome. It is what happens when opaque commercial systems are dropped into public administration without anyone being made responsible for what comes out.
Courts have pushed back, and the EU AI Act sets useful limits. But regulation that arrives after deployment is playing catch-up. What I take from this research is a concern less about any particular AI system than about the general direction of this shift. When the technology that governs people’s lives cannot be closely examined, and responsibility for its decisions dissolves between vendor and state, something essential to democratic life is being lost. Most societies have not yet stopped to decide whether that is a trade they want to make.
