AI governance in the financial sector: clarity, collaboration and compliance

A legal deep dive into the AI Act

After introductory remarks from the ABBL, Aurélia Viémont and Milan Dans of CMS Luxembourg offered a detailed legal overview of the AI Act, outlining its key objectives:

• Harmonising rules for AI systems across the EU
• Safeguarding fundamental rights
• Supporting innovation

They emphasised the Act’s risk-based approach, the classification of AI systems, and the specific governance obligations for high-risk AI applications. Parallels were drawn with the GDPR, offering participants a familiar framework when thinking about compliance.

Bridging research and market: the role of LIST

Dr Francesco Ferrero of the Luxembourg Institute of Science and Technology (LIST) presented ongoing national initiatives designed to reduce the risks and costs of deploying AI in the financial sector.

He introduced the LIST AI Sandboxes, explaining the difference between two key types:

• Technical sandboxes: safe environments for development and assessment using real or realistic data, without impacting production systems or exposing sensitive data.
• Regulatory sandboxes: frameworks that allow experimentation with innovative solutions under temporary regulatory waivers, supervised by a Competent Authority.

Ferrero also showcased tools and services offered by LIST, such as:

• The LLM Social Bias Leaderboard, a freely accessible benchmarking platform
• Tailored AI Assessment Services, designed to test trustworthiness and alignment with ethical AI standards

“By connecting research with market needs, we help financial institutions test and deploy AI responsibly, without compromising innovation,” noted Ferrero.

Governance in practice: insights from the panel

The event concluded with a practical panel discussion featuring:

• Yannick Huchard, Banque Internationale à Luxembourg
• Maxime Hoscheck, European Investment Bank
• Georges Wantz, Deloitte

Panellists shared how their organisations are adapting governance frameworks in anticipation of the AI Act. Key themes included:

• Clarifying internal responsibilities and roles
• Encouraging cross-functional collaboration
• Building internal AI literacy
• Managing the balance between in-house development and outsourcing
• Documenting oversight and using sandbox environments as risk-mitigation tools

The speakers agreed on the need for coordinated efforts between industry, regulators and research institutions, and recognised the AI Act as an opportunity to embed responsible innovation in their AI strategies.