New EU A.I. Directive Proposal from the European Commission

The long-awaited EU Proposal for an #AI Directive was published yesterday. Financial services are not necessarily the primary focus of the Proposal but evaluative uses of AI for credit-scoring and for training or for recruitment purposes are listed as "high risk" in Annex III and will be caught. Certain data-driven law enforcement uses are also deemed "high risk". Robo-advisors will need to inform users that they are interacting with an AI system. The supervisory and reporting framework will be subsumed within CRD4. Uses of AI are divided into risk-sensitive buckets for regulatory purposes: 1) prohibited practices, including: remote biometric real-time identification techniques (eg, facial recognition)--although some law enforcement exceptions apply--and social credit scoring; 2) uses at "high-risk" of contravening fundamental rights, which are subject to certain mandatory requirements and an ex-ante conformity assessment; 3) "interactive" uses--particularly those which are at risk of being manipulative--which are subject to additional transparency requirements; and 4) low risk applications (cf Recital 82). "Artificial intelligence" is defined in terms of outputs and specified by a list of techniques in Annex I. There is provision for an #innovation sandbox.

22 April 2021

1ProposalforaRegulationonaEuropeanapproa.

Download • 910KB

#aigovernance