Requirements and conformity assessment
As mentioned above, we agree that the proposed types of requirements (on training data; data and record-keeping; information to be provided; robustness and accuracy; and human oversight) are important; furthermore, we believe they are relevant for all AI applications, not only those deemed to be high-risk.
We fully agree with the Commission's view that "the conformity assessments for high-risk AI applications should be part of the conformity assessment mechanisms that already exist for a large number of products being placed on the EU's internal market. Where no such existing mechanisms can be relied on, similar mechanisms may need to be established, drawing on best practice and possible input of stakeholders and European standards organisations. Any such new mechanism should be proportionate and non-discriminatory and use transparent and objective criteria in compliance with international obligations."
We agree that "particular account should be taken of the possibility that certain AI systems evolve and learn from experience, which may require repeated assessments over the life-time of the AI systems in question." We note that this could make conformity assessments a significant burden, and we would recommend the Commission considers the potential impact on SMEs in particular and how it could be mitigated.