Answer to the public consultation on Artificial Intelligence

Answer to the public consultation on Artificial Intelligence

On 19 February, the European Commission released a White Paper on Artificial Intelligence (AI); a set of recommendations to foster both an ecosystem of excellence and an ecosystem of trust. A public consultation on this document has been launched on the same date.

Eurosmart answered this consultation to underline that all AI applications – both non-high-risk and high-risk ones- should be covered by mandatory minimum requirements on (1) safety, (2) security, and (3) ethics. Among other things, these requirements should ensure AI cyber-resilience. Security-by-design, including robust algorithms, should be a pre-requisite for all AI applications. Compulsory minimum requirements for all AI applications should also guarantee that AI products and systems are ethical by design, and privacy-friendly. European standards should be used to ensure compliance with such requirements.

For high-risk applications, stricter requirements are needed, as well as mandatory certification supervised by a national authority. Certification should be conducted by duly accredited European laboratories performing their tasks in the EU territory. As AI systems evolve over time, new vulnerabilities may arise. Thus, it is necessary to repeat certification over the lifetime of AI products and systems.

Moreover, Eurosmart believes that AI should be considered a product in the context of the Product Liability Directive. Eurosmart calls on the Commission to adopt a cautious approach when it comes to implementing a strict liability regime coupled with a mandatory insurance for high-risk AI applications. Such an approach should be carefully assessed before being adopted to ensure that it does not discourage manufacturers/users from developing/acquiring AI systems.