Italy in the era of AI: Butti and Floridi propose the idea of ​​the “agentic state”, AI must be governed. The presentation to the Chamber of the Report “Italy in the era of AI. Growth, challenges and prospects of an ongoing revolution”

Floridi needs no introduction (I hope), Butti is the Undersecretary of the Prime Minister with responsibility for Innovation. The report

curated by the President of the Leonardo ETS Foundation, Luciano Floridi, and by Micaela Lovecchio, Education and Training in the Schools of the Leonardo ETS Foundation.

A document that analyzes the national ecosystem and formulates eighteen operational recommendations. The analysis also finds strengths, two supercomputers among the top five in Europe, the first law on artificial intelligence in the EU, alongside several fragilities: dependence on foreign hardware, talent drain and low adoption of artificial intelligence in SMEs.

Excerpts from the interviews, Butti:

“The last two years have seen an acceleration in the adoption of artificial intelligence by businesses. This is accompanied by a structural gap that sees large companies starting to adopt artificial intelligence very seriously, while companies that have 10 employees, or just over 10 employees, reach a maximum of 16% adoption. At Italian level we are at 8.2%. The objective is to reach the European percentage which is 13.5% as soon as possible.. How are we catching up? First of all, we have a strategy that I mentioned earlier, which is a 24-26 artificial intelligence strategy, which then also allowed the approval of a law after a broad parliamentary debate. A law that pays great attention to the world of small and medium-sized businesses.”

“In the three-year plan of information technology for public administration 24-26, which was updated in 2025, we have achieved 150 artificial intelligence projects in public administration*. There’s more, because we aim to* at least 400 in 2026”.

Prof. Florida:

Taking inspiration from the speech of the Hon. Butti, I take up the concept of Agentic state, a nice idea, because it shifts the attention from intelligence itself to action*: systems that do things, that have a concrete impact. But precisely for this reason I am* systems that must be guided, governed*. They are not autonomous entities that act on their own without consequences.*
And here the issue of governance, bioethics, choices immediately comes into play: what is done with these technologies, when and how. On this, it must be said, we are on a good path. And let me say it clearly: this narrative according to which Italy is always and in any case late, always bringing up the rear, is partial*. We remain one of the most relevant countries in the world, and this also entails a responsibility. Because the truth is simple: if something doesn’t happen, it’s because someone didn’t make it happen. If a project does not develop, it is because it was not considered a priority. We cannot always shift the blame onto external factors or a generic system*”

…

A balance is therefore needed: a certain degree of sovereignty and control is fundamental*. We can simplify this by imagining a four-legged table. The first leg is* and give*: we already have important tools on this, such as the GDPR and the AI ​​Act, so there is a solid basis. The second leg is* the software*, i.e. algorithms, models, AI systems. Here the control is much weaker: we have the “ingredients”, but the “cuisine” often doesn’t. The third leg is* the infrastructure*: Where do these systems live? The data centers, the chips, the networks. Here too there is still a lot to do, especially at a European level. The fourth leg is that* normative*, which is one of our strong points*
.…
why should a country spend enormous resources to purchase proprietary solutions, when it could adopt — or contribute to creating, Floridi specified — quality open source solutions*? Of course, today open source is often associated with other models, with other countries, with all the associated unknowns regarding data, governance and ethics. But* why not imagine a European open source, built within the framework of the AI ​​Act and the GDPR*, with clear and guaranteed standards? An open, reliable, certified ecosystem — a “European stamp” — which can be used not only by us, but also by countries such as Canada, Japan, South Korea, Brazil, South Africa and many others*”.