AItocracy and the power behind the algorithm


Jorge Costa Oliveira
A government need never appoint an algorithm to its cabinet for artificial intelligence to acquire political power. It can happen one decision at a time, as agencies use software to screen benefit claims, flag people for investigation, or determine who gets medical attention first. Each decision may look like an administrative detail. Together, they shape how citizens experience the state and how much say they have when it gets something wrong.
Call it AItocracy, a system in which AI acquires a central role in setting public policy, allocating resources, or administering justice. The appeal is easy to see. Algorithms can sift through records in seconds, spot waste in public spending, and help health officials anticipate outbreaks. Consistent rules could curb favoritism and reduce the chance that two people in the same circumstances receive different treatment simply because their applications reach different desks.
In September 2025, Albania presented Diella, an AI assistant, as a virtual minister charged with tackling corruption in public procurement. A software system cannot pocket a bribe. The people who set the tender rules, choose the training data, or control the system can still shape the outcome. A bidder who loses a public contract should be able to see why and challenge the decision before someone with the authority to reverse it.
The same principle applies beyond procurement. A model trained on past lending decisions can reproduce discrimination while presenting its conclusions as objective risk scores. A system that allocates medical resources may look efficient on paper while failing patients whose circumstances are poorly represented in the data. When an official relies on a model whose reasoning is difficult to trace, a citizen may struggle to establish what went wrong, let alone persuade anyone to correct it.
Democratic government derives its authority from the people and depends on their ability to hold decision-makers to account. Deliberation can be slow because citizens disagree about what should count as a good outcome. Extending voting rights, for example, required people to challenge the rules that defined who could participate. A system designed to apply those rules uniformly could have enforced exclusion with consistency.
The concentration of advanced AI in a handful of technology companies compounds the problem. Training leading models requires vast computing resources, data, and specialized expertise. Governments can end up depending on the same vendors to supply the tools, explain their limitations, and advise on their regulation. As those tools become embedded in public agencies, decisions about what to measure and which risks to tolerate can shift from public debate to private design choices. Public scrutiny, independent oversight, and more accessible models can give governments more leverage over what they buy and how they use it.
AI can help officials detect patterns, test assumptions, and explain how they allocate resources. Citizens affected by its use should know when it shaped a decision, be able to challenge the result, and have access to a human official who can remedy an error. That official needs both the authority and the expertise to question the system. Accountability also has to reach the companies that build and maintain it, especially when defects or design choices cause harm.
A government can delegate a calculation. It must answer for the decision, especially when a citizen’s access to care, justice, or a livelihood depends on it.
linkedin.com/in/jorgecostaoliveira
Leave a reply
You must be logged in to post a comment.























