(ANSA) – ROME, DEC 11 – In the end determination prevailed over fatigue.
Europe has made history by delivering an armoury of rules, the first in the world, to regulate the development and use of artificial intelligence systems.
The deal was reached after marathon, 36-hour-long negotiations, giving an idea of the complexity of the challenge the EU took on to draft a judicial framework for AI, weaving together a delicate balance between fundamental rights and support for innovation.
Brussels travelled on unexplored terrain, one that needed to be defined when faced with the opportunities and risks presented by the development of AI, which ‘exploded’ with the spread of chatbots like ChatGPT.
“It’s a historic moment,” said Ursula von der Leyen, celebrating one of the flagship elements of her term at the helm of the European Commission, which presented the proposal in 2021.
“The AI Act will make a substantial contribution to the development of global rules and principles for human-centric AI,” said the head of the EU executive.
This is a reference to the so-called ‘Brussels effect’, with the EU hoping to shape AI rules at the international level, as has happened in other spheres.
“It is a significant step for the development of artificial intelligence in the EU and in Italy,” said Italian Business Minister Adolfo Urso and Alessio Butti, the cabinet undersecretary with the innovation portfolio, in a statement on the agreement, which will have to get final approval from the member States and the European Parliament.
So what does the law create that is new? At the core of the act is the adoption of an approach based on risk.
A series of obligations are envisioned for AI system developers and providers on the basis of the different risk levels identified.
One of the most important chapters, and one which the negotiations were stranded on for hours, regards the AI practices to be banned because they entail unacceptable risks for security and basic rights.
The symbol of this battle is the ban on real-time remote biometric identification systems, such as facial recognition systems, use of which will be limited to specific cases.
“Some governments, including the Italian one, would have liked more freedom in putting citizens under control and profiling but they came up against an insurmountable wall from us in the protection of freedom,” said Brando Benifei, the head of the Democratic Party (PD) delegation in the European Parliament and co-rapporteur of the act.
Amnesty International, however, said the EU has “greenlighted dystopian digital surveillance in the 27 EU Member States, setting a devastating precedent globally concerning artificial intelligence (AI) regulation”.
European Consumer Organization BEUC was also critical, complaining that “the result is underwhelming given the breadth of risks consumers will be improperly protected from in the future”.
One of the most controversial points regards foundation models like GPT-4, which is at the base of ChatGPT.
The agreement foresees more stringent obligations for high-impact models with systemic risks.
These are rules which Berlin, Paris and, to some degree, Rome, would have liked to have watered down to a code of conduct, out of fears the burdens imposed could suffocate innovation in the EU.
“We are not yet convinced that this is the right way to guarantee that Europe remains competitive in AI,” said EPP MEP Axel Voss.
“Innovation will happen anyway, elsewhere.
“Here we missed our chance”. (ANSA).