The European Commission published a proposal for an Artificial Intelligence (AI) Act at the end of April. In parallel, it has set up a public consultation for stakeholders to provide feedback on the draft text, which ended on 6 August.

AI technology has only slowly began arriving on the market and as applications become more sophisticated, they will likely often become very unpredictable in their development. To ensure legal certainty, a level playing field and no obstacles to innovation, a clear definition of artificial intelligence is needed. This would cover the Commission work, as well as national data protection authorities, the Council of Europe initiative, and the OECD framework on classifying AI systems. ESBG members very much welcome the proposed technology-neutral and future-proof definition of AI, and the Commission’s risk-based approach to enable a proportionate regulation.

The Commission aims to turn Europe into the global hub for trustworthy Artificial Intelligence. If we of course share this idea on the principle, it should be recognised that this is a risky bet. Indeed, if European values are not ultimately adopted on an international scale, non-European solutions are potentially more efficient because they have been developed in less restrictive regulatory environments and could compete with European solutions.

With regards to the acceptance of data usage, members would like to use real datasets instead of the Commission proposed ‘synthetic’ datasets. These would mimic real life situations and allow AI training in a realistic setting, without the risk of second order bias (e.g., ethnicity indication based on living area or income).

We also believe that there should be a provision in the draft text to protect European AI developers and users at international level. AI does not discriminate against physical locations, and many different countries across the world have different interpretations of copyright and liability when it comes to AI applications.

Finally, we call for clarity on the scope of the text when it comes to biometric identification of natural persons. It is not yet clear if financial services firms and their providers, who rely on biometric identification to onboard customers remotely (and comply with KYC – know your customer requirements) will be included in the scope of the full set of requirements in the AI regulation.

We support the Commission in its efforts to create a clear legal framework for artificial intelligence which does not inhibit innovation and at the same time provides security for all market participants. We are particularly pleased with the Commission’s philosophical approach to promoting “digitalisation with a human face”. We believe that trustworthy AI in cooperation with human expertise will be of great value to European society. We particularly emphasise the interaction between man and machine. We firmly believe that both humans and machines are irreplaceable. However, we must ensure that new regulation does not inadvertently cripple our markets, dampen innovation and opportunities.