|
Safety, transparency and accountability requirements for AI
MEP Iban Garcia del Blanco (S&D, Spain), rapporteur for the draft report on a framework of ethical aspects of AI, puts forward a proposed Regulation that covers a broad range of aspects.
The proposed Regulation lays down safety, transparency and accountability requirements, including an obligation of robustness. Thus, AI shall be “developed, deployed and used in a resilient manner so that they ensure an adequate level of security, and one that prevents any technical vulnerabilities from being exploited for unfair or unlawful purposes”.
In the case of high-risk AI applications, an assessment of compliance of these technologies with such requirements shall be carried out and monitored by national supervisory authorities.
Iban Garcia del Blanco also recommends creating an EU agency for AI. This agency would help Member States to implement a common approach.
According to Iban Garcia del Blanco, compulsory certification for high-risk applications would co-exist with voluntary certification for other applications. Voluntary certification would take place at EU level and be delivered by the newly created AI agency.
MEPs from other political groups, such as Geoffroy Didier (EPP, France) and Stéphane Séjourné (Renew, France) questioned the need for a new EU agency dedicated to AI.
The issue of consistency between the three reports was raised, especially for the definition of AI and high-risk applications. There are currently discrepancies between the definitions.
Civil liability regime for AI
MEP Axel Voss (EPP, Germany), rapporteur for the draft report on a civil liability regime for AI, insisted on the need for a full harmonisation to create a functioning Digital Single Market. He does not want completely new liability rules but prefers adapting what already exists. Amendments to the Product Liability Directive might be needed. Axel Voss underlined that new rules should not hamper the development of AI.
In his draft report (including a proposed Regulation), Axel Voss envisages a strict liability regime for high-risk AI systems, coupled with mandatory insurance. Other AI systems would be covered by fault-based liability.
Axel Voss’s definition of deployers and his choice to focus on them for liability was questioned during the debates.
For more information on this report, see previous briefing here.
Next steps:
25 May: deadline for amendments on the reports
Summer (late June or early July): amendments to the reports are expected to be debated in the JURI Committee
28 September: committee vote on the reports
October: plenary vote
For any question on this issue, do not hesitate to contact Camille Dornier: camille.dornier@eurosmart.com
|
