Lhe EU has been working for years on a draft regulation on artificial intelligence (AI) which would be the first major legislation in the world to regulate the sector. With the objective of placing the Old Continent at the head of the race for innovation while guaranteeing the security and rights of users.
The Commission presented its draft in April 2021. MEPs hope to adopt their position on the text at the end of March, paving the way for a negotiation with the 27 member states which should take another months. European Commissioner for the Internal Market, Thierry Breton, explained that the three EU institutions were already working together to “clarify the rules further” on technologies such as ChatGPT.
“We need a strong regulatory framework”
Created by the Californian company OpenAI, this editorial content generator has made headlines since its revelation in November. In a few seconds, he can answer complex questions clearly and in several languages, write sonnets or write a scientific article. Its creators claim that the latest version, GPT-4, even passed the bar exam.
This power is not without danger. If the teachers fear that the pupils use it to cheat, the political decision-makers fear especially that it could be used for misinformation. “As ChatGPT shows, AI solutions can provide great opportunities for businesses and citizens, but they can also present risks. This is why we need a strong regulatory framework”said Thierry Breton.
ChatGPT is just one example of the potential of so-called general-purpose artificial intelligences, i.e. capable of performing a wide variety of tasks, including generative AIs that can create original content such as images and text. , from large amounts of existing data.
“We will propose a set of rules to govern general purpose AI”, MEP Dragos Tudorache, one of the text’s co-rapporteurs, told AFP. Together with fellow co-rapporteur Brando Benifei, he presented a paper last week proposing obligations for general-purpose artificial intelligences, which were not included in the Commission’s initial proposal.
Experts had warned of the risks associated with ChatGPT-type systems, and briefed EU officials early on in the draft regulations. “Our recommendation at the time was that we should also regulate AI systems that have a wide range of uses”, said Kris Shrishak, head of technology at the Irish Council for Civil Liberties. He stressed that it was equally important to identify the risks associated with generative AI systems after they were deployed.
For Kris Shrishak, the effectiveness of the new regulation will depend on the final text, but the draft “represents a good basis and contains mechanisms to identify new risks”.
The most crucial issue, however, will be the ability to enforce the law, he warned, adding that the European Parliament was working to strengthen this aspect. “Rules are just a piece of paper if not enforced”he told AFP.
Last week, OpenAI CEO Sam Altman proposed that “major governments of the world” and the “trusted international institutions” come together to produce a set of rules specifying what the system may or may not do.
As for how Brussels should regulate AI, the content generator believes that “the EU could take a step-by-step regulatory approach, building on ethical governance principles, certification and assessment standards, appropriate oversight and international collaboration”.
(with AFP)