EU Proposed AI Legal Framework

9 Pages Posted: 18 May 2021

See all articles by Emre Kazim

Emre Kazim

University College London ; Holistic AI

Charles Kerrigan

CMS London

Adriano Koshiyama

Department of Computer Science, University College London

Date Written: May 15, 2021


The publication of the EU’s draft AI legal framework is a milestone in the regulatory debate on AI. It proposes a risk based approach to regulating and reporting. In this white paper, we provide a high-level overview of the risk tiers, which we take to be the kernel of the legislation, and follow this by offering our initial thoughts and feedback on strategic points of contention in the legislation. Our main takeaways are: (i) Innovation - the sandbox approach may not be enough to ensure innovation; (ii) Reporting - in the lead up to codification we would like to see reporting being used to accelerate dissemination of best practice and benchmarking; (iii) Green-flagging - there does not appear to be sufficient detail to derive a reasonable set of green-flagging conditions; and, (iv) Manipulation - addressing the ambiguity in the draft proposal of banning systems with ‘significant manipulation’. We conclude with notes on the legal status of algorithms, the status of GDPR in light of AI regulation and, the geopolitical ramifications of EU AI regulation.

Keywords: Regulation, Compliance, Artificial Intelligence, Accountability, Governance, Fairness, Transparency, European Union

Suggested Citation

Kazim, Emre and Kerrigan, Charles and Koshiyama, Adriano, EU Proposed AI Legal Framework (May 15, 2021). Available at SSRN: or

Emre Kazim (Contact Author)

University College London ( email )

United Kingdom

Holistic AI ( email )

18 Soho Square
London, W1D 3QH

Charles Kerrigan

CMS London ( email )

United Kingdom

Adriano Koshiyama

Department of Computer Science, University College London ( email )

Gower Street
London, London WC1E 6BT
United Kingdom

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics