More on the AIR draft
by Guido Scorza, member of The Italian Data Protection Authority
He considers that, as with #privacy legislation some years ago, it is a great initiative beyond any reasonable doubt
Some + points
– Regulation instead of a Directive (+uniformity)
– Importance of #datasets used to train models to avoid inaccurate or poisoned data compromised the outcomes
– introduction of sandboxes to boost innovation
Some doubts
– the idea of including a fixed -albeit subject to amendments by the EC- list of high-risk #AI systems and forbidden practices in a field where rapid innovation is the rule [true, but it is a sensible starting point provided the EC actively updates it]
– the approach to the problem of #transparency: imposing operators a duty to inform individuals about the AI systems when nobody reads long lists of T&C or privacy notices [true, but it holds providers accountable. Maybe the inclusion of a label -like Apple- would be + helpful for individuals]
– AI governance and whether we need another EU institution (EAIB) and a new network of national competent authorities [while the AIR draft establishes the new EAIB, the new supervisory functions at nat level might be carried out by existing authorities, e.g. DPAs, since it is not ruled out, art59 AIRd]
Le regole Ue sull’intelligenza artificiale – Intervento di Guido Scorza – MF
0 comentarios