The House approves the AI ​​bill. Here's what it provides

The Chamber of Deputies has given the green light to the bill (ddl) that delegates the Government to regulate the use of artificial intelligence in Italy. The measure now returns to the Senate for a third reading, but it already represents a crucial step in the national regulatory framework: Italy could become the first country in the European Union to adopt a comprehensive national law on AI, implementing and integrating the AI ​​Act, the European law on Artificial Intelligence.

What does the IA bill provide?

The text regulates the use of artificial intelligence in the public and private sectors, with a focus on professions, work, justice, public administration and copyright protection.

Among the key points: obligation for professionals (such as lawyers and accountants) to inform clients on the use of AI; prohibition for employers to use artificial intelligence tools for remote monitoring of employees; clear limits on the so-called "predictive justice", admitting AI only for technical or organizational support to the activity of magistrates.

Butti: “Fundamental step for Italy”. Ascani: “We will be colonies of empires”

Undersecretary for Innovation Alessio Butti was satisfied and defined the approval as “a fundamental step for Italy”, adding: “I hope for a quick third reading in the Senate that will lead us to officially be the first European country to have a dedicated national law. A result that is by no means a foregone conclusion, due to the strong will of the Government and the work shared with all parliamentary groups. After having guided the drafting of the European AI Act and the international G7 principles, we are now completing the national framework, strengthening security, innovation and protection of rights”.

On the other hand, the comment of the vice president of the Chamber Anna Ascani (PD) was critical, who, while recognizing the work done in the Commission, defined the text as "a missed opportunity to affirm a true European technological sovereignty". "The measure could have been the tool to fight the new AI oligarchies, promoting investments in European proprietary technologies and sovereign cloud. Instead - she declared - our amendments on algorithmic transparency and model training were rejected. A serious and inconceivable choice. Technology is not neutral, and without democratization tools, it risks becoming a powerful instrument of domination".

Then the thrust: "If a European way is not found, if the resources of the twenty-seven member countries are not pooled, we will only be able to choose which global empire we want to be colonies of". The Network for Digital Human Rights also criticizes, denouncing the lack of democratic checks and the excessive power entrusted to the Government. Proposals on independent authority, transparency and limits to biometric surveillance were rejected. "A missed opportunity to protect fundamental rights", the Network comments.

The Article 8 issue: what changes for research

The bill also provides that, where possible, the technologies adopted by public administrations and professionals must be Italian or European, while the possibility of storing public data on non-EU servers has been confirmed. Delegations to the Government include professional training, definition of fair compensation in relation to the use of AI and supervisory mechanisms.

An important novelty arrives for research. Article 8 of the bill represents a novelty in the field of personal data management for research on artificial intelligence and related sectors, such as robotics, neuroscience and brain-computer interfaces.

Based on an exemption provided by the GDPR, it qualifies such processing as being of “significant public interest”, thus allowing it to proceed without the explicit consent of the interested parties, provided that the subjects involved are public bodies, private non-profit organizations or IRCCS engaged in research projects.

Furthermore, the article provides that the Privacy Guarantor does not have to provide preventive authorizations, but can only actively intervene to block any non-compliant treatments. Finally, the rule clarifies that the approval of the ethics committees is mandatory only for the processing of identifiable personal data, while it is not required in the case of anonymized data, thus favoring greater agility in scientific research, although raising questions on the balance between privacy protection and innovation.

The third reading in the Senate is expected in the coming weeks.