2023-12-09 05:40:33
White smoke in Brussels. After tough negotiations and difficult discussions between the European Parliament and member states, the European Union has managed to agree on a historic text to regulate artificial intelligence (AI). Historic because it de facto makes the EU the first continent in the world to adopt a complete legal framework.
The negotiators “topped” after two long sessions of trilogue negotiations, final discussions between the Parliament, the Council, and the Commission, a marathon of more than 35 hours. “Finding the balance between user security and start-up innovation, while respecting fundamental rights and European values, was not easy, but we succeeded,” welcomed the European Digital Commissioner in a press release. , Thierry Breton, at the origin of the legislative proposal presented… in April 2021!
Historic!
The EU becomes the very first continent to set clear rules for the use of AI ????????
The #AIAct is much more than a rulebook — it’s a launchpad for EU startups and researchers to lead the global AI race.
The best is yet to come! ???? pic.twitter.com/W9rths31MU
— Thierry Breton (@ThierryBreton) December 8, 2023
At the time, ChatGPT did not yet exist, and it was therefore necessary to integrate into the negotiations this new family – which has continued to grow for a year – of generative AI capable of generating on command, and with astonishing relevance, text, sound, and image. Capable of competing with humans on intellectual tasks, for the first time, these applications were also quickly singled out for their ability to generate false content, larger than life.
Regulate… without restricting yourself
Concerned about limiting the excesses of AI, MEPs were the first to ask for specific supervision of these technologies, while Member States were especially concerned not to restrict innovation at a time when local champions are in the process of hatch – Aleph Alpha in Germany or Mistal AI in France.
Ultimately, the agreement plans to require GPT-type models to have a minimum of transparency on the algorithms and databases that make them work, with additional restrictions (human control, risk management program) only being planned for the models. the most elaborate – deemed high risk – essentially those used in sensitive areas such as critical infrastructure, education, human resources, maintaining order, etc. The strongest barrier, the ban, would only be reserved to tools “contrary to EU values”, like citizen rating systems or mass surveillance used in China.
Finally, on one of the most divisive points, the remote biometric identification of people in public places by the police, the States and Parliament seem to have found a compromise, with a ban in principle reaffirmed but accompanied by a “small number” of exemptions. “We do not want mass surveillance in Europe,” however, insisted Thierry Breton in his press release, also announcing the creation of a European AI monitoring office, and significant penalties for companies that wish to contravene the new rules.
#historic #European #agreement #regulate #Artificial #Intelligence