Artificial intelligence, the time has come for operability

It is time to do. Artificial intelligence is talked about on a daily basis, and critical issues and opportunities are analysed, but it is time to move on to using artificial intelligence in companies and public administrations. It is certain that no one has the solution ready, but we must experiment in order not to fall behind. Our companies have already started and have already moved beyond the stage of fear, so much so that - as emerges from the report on AI by Aspen Institute Italia and Intesa San Paolo - 58% have already adopted it for basic tasks such as predictive data analysis, customer service automation via chatbots and to increase IT security. And in the very near future even 98 per cent of companies say they plan to use it. The call for concreteness could not fail to come also from the President of Confindustria Emilia, Valter Caiumi, who, at the assembly a few days ago, emphasised that 'at the end of 2022, the AI market in Italy was worth 435 million euros. The budget of Germany and France on AI is 2 billion each. These are unreachable figures for us. So, as on other occasions, we have to focus on adoption: the doing'.

Also in Cernobbio, the research 'AI 4 Italy: from theory to practice', elaborated by TEHA Group in collaboration with Microsoft Italia, underlines that 100 per cent of Italian companies use or intend to use generative AI in the short to medium term, that no Italian company declares itself not interested in generative AI solutions, and confirms that AI is key for the Made in Italy.

The Italian Artificial Intelligence Strategy 2024-26 emphasises the 'risk of not doing' while respecting fundamental rights.

So, do, but how? At this stage, one can only experiment, also from a legal point of view. And therefore, as many insurance companies and companies in the financial sector, some banks, some publishing houses, and some public bodies and healthcare institutions have already done, give themselves internal policies or codes of conduct that serve to clearly indicate what can and cannot be done, who is responsible within the organisation, who must supervise, what information must be provided to the customer or user, and so on. Crucial is the contract with the supplier of the customised AI system, and the data ownership clauses deserve special attention. Indeed, it is not only the software to manage artificial intelligence that is relevant, but the personal and non-personal data that interacts with it makes all the difference. The ownership of the outputs generated by the AI system must be contractually defined, which is not necessarily either of the person who developed the software or of the person who contributed the data. Another clause to be carefully considered is that of the usability of the data for training the programme. Upstream, of course, the physical availability of the data must correspond to its legal availability. In other words, it must be verified that personal data or non-personal data required for AI programmes can be legitimately available.

The challenges for Italy, indicated in all the Reports, are focused on strengthening investments and finding AI solutions for the 'made in Italy', which enhance the Italian heritage, and first of all, the Italian cultural heritage and support Italian companies, such as those in manufacturing. This is combined with the need to invest in building digital competences on AI.