The integration of Artificial Intelligence (AI) into the public sector is being promoted globally as the "holy grail" of administrative efficiency. From automating tax audits to managing social welfare benefits, the promise is clear: faster services, reduced bureaucracy, and cost savings. However, a recent report highlighted by Oikonomikos Tachydromos rings the alarm about the "uncontrolled" uses of AI, which often develop within an institutional vacuum, far from public scrutiny and without the necessary accountability.

The Greek Reality and Digital Transformation

In Greece, the effort to digitize the state over recent years has been impressive. Gov.gr has changed the daily lives of citizens, but the next phase—that of "smart" administration—carries risks that have not been sufficiently analyzed. The use of algorithms by the Independent Authority for Public Revenue (AADE) to detect tax evasion or the implementation of predictive systems in employment services (DYPA) are examples where AI makes decisions that directly affect citizens' lives and property. The problem lies in the fact that many of these algorithms operate as "black boxes," where neither the citizen nor often the public official knows the exact criteria upon which a decision was made.

  • Lack of transparency in code and training data.
  • Risk of embedding biases that lead to systemic discrimination.
  • Absence of a clear legislative framework for appealing algorithmic decisions.

The Accountability Deficit and the Risk of Discrimination

International experience has shown that blind trust in algorithms can be catastrophic. In the Netherlands, the SyRI scandal, where an AI system wrongly accused thousands of families of benefit fraud, led to the resignation of the government. In Greece, the risk is similar if there is no strict oversight. When the state delegates decision-making to software, the concept of administrative justice is shaken. How can a citizen defend themselves against a decision based on statistical probabilities rather than tangible evidence?

"Technology is not neutral; it reflects the biases of its creators and the data that feeds it. In the public sector, this can translate into digital authoritarianism."

The Challenge of AI Act Compliance

Despite the passage of the European AI Act, its implementation in the public sector remains a challenge. Public authorities are often exempted or enjoy greater freedom under the pretext of "public safety" or "national necessity." There is an urgent need for the creation of independent oversight bodies to conduct regular audits of public sector AI systems, ensuring that technology serves humanity and not the other way around. Digital convergence must not come at the expense of democratic achievements.

Conclusion: Towards Ethical Digitization

Artificial Intelligence can indeed modernize the state, provided it is not used as a tool for evading responsibility. Transparency, explainability, and human control (human-in-the-loop) must be the core principles of every new application. The challenge for Greece in 2026 is to prove that digital progress can coexist with the protection of fundamental rights, avoiding the pitfalls of an uncontrolled algorithmic state.