AI oversight is a board responsibility.
I am Katia Ciesielska, a Luxembourg independent non-executive director. Not a technology project delegated downwards. The EU AI Act creates obligations across the organisations and service providers that fund boards oversee. The first challenge is knowing where AI is actually being used, and who owns the resulting obligations.
Most fund boards look for AI in the wrong place.
A Luxembourg fund may have no employees and no systems of its own. The AI is at the delegates: the AIFM, the central administrator, the transfer agent, the AML screening vendor. It arrives as a feature of an existing service rather than as a new one, which is why it rarely reaches the board.
In my experience AI appears in board packs as an innovation topic, and almost never in the section where the board reviews its service providers.
What AI is embedded in the services you deliver to us, and how are the AI Act obligations allocated between us?
Deployment has outrun oversight.
Seven commitments set out in AI in the Boardroom, the ILA guidance I co-authored with the Research Think Tank.
To govern AI, boards must elevate.
What the board should require, in what order. Visibility before controls, controls before scale.
The first ninety days.
Published and presented.
Co-author of the ILA’s board guidance on AI, and a regular speaker on AI governance and EU AI Act readiness across Luxembourg and Brussels.
What does AI governance mean for a Luxembourg fund board?
Fund boards in Luxembourg face a two-layer challenge. The first is operational: AI is already embedded in services most boards delegate — AML screening, transfer-agent processes, administrator workflows — and those systems fall within the board’s oversight of its delegates under AIFMD. The second is regulatory: since 2 August 2026, EU AI Act transparency obligations apply to providers of AI systems, and boards with CSSF-regulated service providers need to understand whether those obligations affect their own governance arrangements. Effective AI governance at board level means asking the right questions of service providers, not deploying AI internally.
Is an independent director with AI governance expertise different from other iNEDs?
The combination is still rare in Luxembourg. Most directors either have fund-governance experience or familiarity with technology; very few have both, with published board-level work on the topic. For boards that need to demonstrate that their oversight of AI risk is substantive — to regulators, to investors, or in the context of DORA — a director who can engage at that level rather than simply approve a policy adds something different.
What is the ELEVATE framework?
ELEVATE is a board-level AI governance model developed in the ILA publication AI in the Boardroom: Opportunities, Risk and Board Oversight (2026), co-authored by Katia Ciesielska. It stands for Educate, Legislate internally, Evaluate, Verify, Anchor human judgment, Test and adapt, and Ensure safety. It is designed to give boards a structured approach to AI oversight that goes beyond a checklist — connecting regulatory compliance, risk management and strategic judgment.
Looking for an independent perspective?
For board appointments, governance advisory or speaking enquiries, I would be pleased to hear from you.
KATIA@KATIACIESIELSKA.COM →