It was a very long night among the negotiators of the European institutions who were preparing to finalize the artificial intelligence law. After almost 22 hours without stopping, the discussions between the Commission, the European Parliament and the EU Council, which represents the States, were concluded on Thursday and will resume today.
The negotiators arrived ready to close one of the key laws of this legislature but, as expected, the discussions were not going to be easy. “A lot of progress has been made in the last 22 hours,” said the European Commissioner for the Internal Market, Thierry Breton, on the social network artificial intelligence based on the risk it entails.
“We know that it is not going to be easy, that it is not going to be resolved in a couple of hours (…) but the alternative of not reaching an agreement is the status quo and right now there is no control over anything,” sources in the negotiation warned. before starting discussions. During the night progress was made on important issues. One of the main ones is how the use of what are known as foundational models, such as ChatGPT, is regulated, although the details are unknown. When the European Commission proposed its law in 2021, then this technology did not even exist. Now, it is in its fourth version. For this reason, the European Parliament insisted that its regulation should be introduced into the law, even more so after the impact it has had on the sector and to which other systems have been added, such as Bard or the recent launch of Gemini (both from Google).
Shortly before yesterday’s meeting began, positions were far apart. The European Parliament required developers to show that foundational models are safe, especially when used in risky contexts; or that the system does not discriminate based on race. Instead, countries opted for self-regulation by the companies themselves.
But as expected, the main point of friction is in what context a technology should be banned or considered high risk. Negotiators were engaged in where biometric surveillance, such as facial recognition cameras, can be accepted. For the European Parliament, it should be banned completely, although the European Parliament’s negotiator, the social democrat Brando Benifei, was open to allowing it in some cases and with safeguards; while states consider that it can be used in very specific situations, such as in cases of terrorism or national security, – especially important for France – and under judicial authorization.
The negotiators hope to finalize today one of the key laws for Spain, which holds the rotating presidency of the Council. Even the first vice president, Nadia Calviño, taking advantage of the fact that she was meeting with the Eurozone Economy Ministers, approached the meeting room to inquire about the status of the discussions. “Stay alert,” warned Commissioner Thierry Breton.