The European Parliament was due to vote on Tuesday 20 October on a set of key principles to frame artificial intelligence and guide future proposals from the European Commission. However, at the time of going to press, the results of the final vote were not yet known.
The Commission’s projects
In its work programme, the European Commission calls for “legislation covering security, liability, fundamental rights and data aspects of artificial intelligence”.
Although this initiative does not appear in any of the annexes to the 2021 work programme, Commissioner for Justice Didier Reynders confirmed the Commission’s intention to legislate on this issue as early as next year at an online seminar organised on 14 October by the Centre on Regulation in Europe (CERRE).
In line with the February White Paper, he referred in particular to the idea that a presentation document or a test would be required, before placing an artificial intelligence product on the market (see EUROPE 12429/5). He also raised the issue of hazardous applications and their opacity. And it is precisely these issues that MEPs raise in their resolutions.
Three separate reports
The own-initiative legislative report by Ibán García Del Blanco (S&D, Spain) on ethical aspects defends human-centred artificial intelligence. Supporting the idea of ex ante control on the most high-risk applications, he proposes establishing “an exhaustive and cumulative list of high-risk sectors and high-risk uses or purposes”. The amendment introduced by the GUE-NGL group in favour of a moratorium on facial recognition in public open spaces until the risks associated with it have been eliminated was rejected. [https://bit.ly/3dJCp21 ]
The own-initiative report drafted by Axel Voss (EPP, Germany) supports a compulsory liability regime for high-risk artificial intelligence systems, as well as compensation measures in the event of damage. “Our economic survival is at stake”, said the rapporteur before the vote. “Compensation must be available, regardless of whether it is the algorithm, software or hardware that is at fault”, he said. [https://bit.ly/2TcpOv2 ]
Finally, the non-legislative own-initiative report drafted by Stéphane Séjourné (Renew Europe, France) on intellectual property rights for the development of AI-related technologies opposes any desire to recognise legal liability for an AI. The text recommends that a distinction be made between production “aided by AI” and production “entirely generated by AI” and emphasises that where AI is used solely as a tool to assist an author in the creation process, the current copyright framework remains applicable. In the vote, however, MEPs opposed extending the obligation to keep verifiable registers to data containing images and/or videos with biometric data. [https://bit.ly/34gD2wZ ] (Original version in French by Sophie Petitjean)