Artificial intelligence in the public sector: Speed or trust?

I previously wrote in this column that the artificial intelligence debate is not just a technology race, but that the real issue is trust, governance, and public responsibility. I mentioned that with the wave of regulation in Europe, especially with the European Union AI Act, governments have ended the era of “act first, think later.” I stated that while artificial intelligence is spreading rapidly in the public sector, when it comes to the state, trust comes before speed.

The assessments shared during the SAS event I attended this week are essentially a continuation of that article. As we head into 2026, investments in artificial intelligence in public institutions are increasing. Generative AI solutions, automation systems, and even software “agents” capable of making decisions on their own are entering the field. However, there is one area that is not growing at the same speed: the infrastructure of reliability.

For the public sector, the issue is no longer “more artificial intelligence.” It is about more transparent, more auditable, and more accountable artificial intelligence. Because a decision made by the state is not limited to a credit card limit; it involves issues that directly affect lives, such as social assistance, tax audits, and health surveillance.

What is RAG, and why is it important?

One of the concepts frequently mentioned at the event was RAG. The technical expansion is: “Retrieval Augmented Generation.” Let me explain it simply. Instead of an artificial intelligence model answering only with general information it learned from the internet, it means producing answers by looking at the institution's own archives, legislation, and past decisions.

In other words, the system does not speak off the cuff; it relies on institutional memory. This is critical, especially for the public sector. Because the state works with memory. Thanks to RAG systems, years of accumulated knowledge are being digitized, and it is effectively creating an “artificial intelligence mentor” for young employees.

There is also a risk here. If institutional information is not transferred correctly, the system learns incorrectly. The issue of trust begins exactly here.

Sovereign AI and digital borders

Another topic highlighted by SAS experts is “Sovereign AI.” Countries' desire to keep data and computing power within their own borders is increasing. Data centers, national models, and local infrastructure are becoming strategic.

This is not just a technical issue; it is a geopolitical one. Data is no longer oil; it is like the refinery that manages the oil. Whoever processes it, that is where the value is created.

Public servant or expensive consultancy?

Another notable trend: States prefer to empower their own employees rather than massive consultancy projects. The goal is for the public official to do more work with fewer resources using tools that accelerate analysis.

Technology alone is not the solution. Unless it is combined with human competence, it will just be expensive software.

A new era in tax, health, and social assistance

Artificial intelligence is making fraud networks more sophisticated. Fake identities, automated content generation, complex financial manipulations. The state is responding to this with artificial intelligence as well. Thanks to real-time analysis, instant warning systems, and inter-institutional data sharing, the goal of reducing the tax gap is strengthening.

On the health side, digitizing data that still sits in paper files is a revolution in itself. When these documents are interpreted with artificial intelligence, epidemic detection speeds up, reporting is simplified, and duplicate records are reduced.

The real question of 2026

It is possible to summarize the whole picture in one sentence:

2026 will be the year when artificial intelligence in the public sector shifts from quantity to quality.

The question is: Will states focus on the speed race, or will they build a trust architecture?

My opinion is clear. Success in the public sector will not be measured by building the largest model, but by building the most reliable system.

Technology evolves. Models change.

But the reputation of the state, once damaged, is not easily recovered.