Benifei (Pd): “On Ai and facial recognition, the EU will be a model. We will oppose Piantedosi’s security line”

Europe could adopt the first law in the world that regulates artificial intelligence. What is the purpose of this regulation?

“The European Union wants to establish guarantees in favor of users of those services and products that include artificial intelligence, especially when a certain level of risk to health, safety and fundamental rights is recognized. The goal is precisely to prevent and mitigate risks, through certification procedures on data governance, human control, technical specifications and more, increasing the potential for consumers and businesses, to create development and opportunities. Finally, it can be said that we expect to obtain a certain ‘Brussels Effect’: in the rest of the world there is already a debate around our rules and we hope that our regulatory model can be a basis for comparison with other countries, for this reason we have already carried out a dialogue within the OECD and with other international organizations.

On which aspects of the law do you think there will be more political confrontation in next month’s plenary?

“I am thinking in particular of two proposals by the parliament, which were not present in the draft of the Commission. The very strong limitation of biometric recognition by the police forces for the surveillance of public places, which in the draft included too extensive exceptions. And the forecast an impact assessment on fundamental rights for high-risk AI. In general, we have also increased the number of prohibited use cases, for example some areas of emotional recognition, to ensure greater protection of people. This list could be subject of debate but in reality, given the large majority of this vote, I hope that we will proceed without too many problems to approve the text in the plenary. The real challenge will be the subsequent confrontation with the governments on some of these more delicate points”.

The text also tries to regulate social scoring, facial recognition, predictive policing. What relationship between society and technology are you trying to achieve?

“A relationship of trust rather than fear. AI must not become a control tool and we do not want to allow its use, perhaps initially for a good purpose, to turn into abuse. Unfortunately, history teaches us that once that the door is open, although there are exceptions, it will always be difficult to verify them one by one. This is why we don’t want to take risks and instead support all the wide positive potential of this technology to the maximum, from school to work to business”.

How does Europe try to position itself on the AI ​​theme compared to China, the USA and their companies?

“We certainly don’t want to wait for the worst to happen before having rules. And then we must not forget that there are cases, even in Europe, in which AI has been used to the detriment of people. We have seen them in Italy, in Holland, in the United Kingdom and made us reflect on what was necessary.The truth is that now even China and the United States, especially with the acceleration of recent months due to the diffusion of large language models among millions of users, are ran for cover. The difference is that the European Union has been working on it for many years”.

ChatGpt has sparked a strong debate on the role of AI. Among the most felt risks is that disinformation could be created through these tools, with damage to the stability of democracies. How is this problem addressed?

“We have new requirements for generative AIs like ChatGpt. They will need to transparently declare each time a text is generated by the AI, as well as provide a detailed summary of the copyrighted training data. These choices will allow the reader to know that is reading something not written by a human being (and therefore subject to a higher percentage of error and inaccuracy) and will also facilitate the requests of those who produced the material on which the AI ​​trained, with possible compensation, where Furthermore, the fight against the production of illegal content will be facilitated, as well as the identifiability of deepfakes”.

Another fear is the impact on the world of work.

“As far as jobs are concerned, it will be essential that the European Union and governments maintain an adequate commitment in terms of investments, training, the choice to carry out common projects, in order to effectively manage the transition and increase job opportunities in the various fields development and control of AI”.

Still on the subject of work, there is the subject of the control of workers through artificial intelligence systems.

“The control of workers will be limited, in addition to the GDPR, by the ban on biometric recognition, emotions and social scoring. The need to control productivity must not lead to surveillance and discrimination of workers and for this we have inserted a clear safeguard on the need to involve the representatives of the social partners to introduce AI in the workplace and we have clarified the possibility of developing more protective rules for workers without this being contestable as divergent from the regulation which remains an internal market regulation. I have fought a lot on these issues by making them a priority for my political group because I don’t want these technologies to be used to widen inequalities and power imbalances”.

Have there been any meetings between European legislators and top management of companies like OpenAi in this period?

“Institutions have been meeting stakeholders for years, from businesses to civil society. It is a normal and useful process, both in the drafting phase by the Commission and in the revision phase by the Parliament and the Council. Definitely they will also be there in view of the trialogue and will be useful to better understand what the co-legislators still have to overlook. In this regard, it is very important to have maximum transparency on who has met and also to discuss frankly the different approaches in the field”.

One of the provisions of the text concerns the application of facial recognition in public spaces. Minister Piantedosi recently declared his intention to strengthen it in some critical places.

“The only positive thing I can say is that I am pleased that the Guarantor is already working to ensure that a possible proposal, which in any case does not see my favor, respects the fundamental right of privacy. In the past we have seen municipalities and law enforcement agencies install these systems without respecting those guarantees that European legislation has been proposing for some time and that we are trying to strengthen with the AI ​​Act in the field of biometric recognition.The approach is not based on data but prejudicially security of ministers like Piantedosi and other European ministers will be firmly opposed because we don’t want a Europe at risk of “Big Brother” and given the power of these systems we don’t think this is an exaggerated ‘slogan'”.

Would it be possible to do this with this standard?

“If it were up to me it should be banned in a total way. As Parliament at the moment the common position is that biometric recognition in real time is prohibited and there is a very limited possibility of use by the judicial authority only in case of deferred use and for crimes already accomplished. There are so many errors of this technology that the risk of stopping innocents starting from confused and undefined images is too high. Fortunately, at the moment there is a moratorium in Italy pushed by the Democratic Party and in particular by the parliamentarian Filippo Sensi “.

What impact did the Guarantor’s decision to stop ChatGpt in Italy have?

“The intervention of the Guarantor made everyone raise their antennas, not only in Europe, I was in the United States in the days of the decision and there was a lot of talk about it. Many other authorities have started their own investigations and in any case if today we are thinking about this new regulation, we have a duty to enforce the existing rules”.

How soon will Europe be able to adopt this law?

“In mid-June we will vote on the text in the plenary session in Strasbourg and then we will start the negotiation trilogues with the Commission and the Council. We want to conclude within the year in order to approve the standard definitively at the beginning of 2024, the application of which will begin gradually for successive steps and will require a total of almost two years. I am convinced that there will still be discussions on the possibility of further accelerating the entry into force, which I am in favor of”.