Evaluating Responsibilities Across the Value Chain: AI Act and its Consequences for European SMEs

By Triinu Viires

On Tuesday November 15th, SME Europe organized the Webinar “Evaluating Responsibilities Across the Value Chain: AI Act and its Consequences for European SMEs” on the initiative of Kai ZENNER, Digital Policy Adviser & Head of Office Axel Voss MEP.

The panel was opened by Michal BONI, Member of the European Parliament 2014-2019; First Minister of Administration and Digitalisation of Poland; Senator of SME Europe of the EPP; and saw the participation of Kai ZENNER; Alex ENGLER, Fellow, Brookings Institution; Associate Fellow, CEPS; Giorgos VERDI, Junior Policy Officer, Digital SME Alliance; Alexandra BELIAS, International Public Policy Manager, Deepmind. The discussion was moderated by Eline CHIVOT, Senior Adviser on Digital Policy, EPP.

Michal Boni opened the event stating that after nearly three-four years of the debate on artificial intelligence development based on AI strategy and proposed on the spring 2021 AI Act, we are much more advanced in understanding what kind of regulatory framework is needed, but also what are the gaps in the legislative proposal. What is still not clear – the unintended consequences of the artificial intelligence the AI Act can bring and what kind of impact it has on many areas functioning with artificial intelligence systems , especially on SMEs.

Kai Zenner said that the Commission envisaged that a strong focus of the AI actors should be on the provider that needs to fulfill high risk obligations from article 8 to 15 and that they would also need to do a lot of monitoring cooperation with market surveillance authorities. He added that Art 28 in the Commission’s proposal is not adequately addressing the complex Al supply chain with often more than one provider/user. Do they share the Al Act obligations or is there one ‘lead’ actor? What happens if an Al system that was placed on the market/put into service becomes high-risk only after a modification? Certain obligations for providers/users are missing or should be specified or redistributed.

Alex Engler adds that one of the first things you can do is clarify who do you expect the provider to be. He suggests to not regulate open-source GPAI as it has many downsides and few upsides and rather really focus on getting AI Act to work in a value chain. He stresses to not confuse EU AI Act main point – commercialization risks – with proliferation (disinformation, deep fakes), for which a different policy recommendation is needed.

Giorgos Verdi remarked that from the market viewpoint, the foundation models are characterized by intense amounts of capital (large amounts of computational power and data to develop) and that most small and medium enterprises lack these resources. Recent legislative developments were seen generally as positive by Digital SME Alliance member base. He adds that the AI ecosystem is a very competitive place so it is important that the rules of the game are clear so that European companies are able to compete and accelerate on the uptake of AI.

Alexandra Belias concluded that we need to foster culture of exchanging best practices and raise the bar constantly in the EU of what it means to have safe ethical AI.