Artificial intelligence: uses and FINMA’s expectations

Read the PDF edition (3 pages)

Generative artificial intelligence made its entry into Geneva’s institutions from 2023 onwards, first through individual uses and then through organised projects: document summaries, assisted drafting, translation, information extraction, support for advisers, transaction monitoring, portfolio analysis and client responses. The most advanced uses concern compliance, where name-screening tools, transaction analysis and open-source research reduce the number of false alerts and the time spent on analysis. They also concern commodity trading, where forecasting models incorporate ever larger volumes of data.

FINMA set out its expectations in a supervisory communication of 18 December 2024 on governance and risk management in connection with the use of artificial intelligence. It expects supervised institutions to have clear governance, an inventory of applications, a risk classification, assurance of data quality, testing and validation, documentation, mechanisms of explainability and independent reviews. It noted that most institutions were at the early stages and that the corresponding structures were still being built.¹ The risks identified include model risks, a lack of robustness, accuracy or explainability, or bias, data security, cyber risks, dependence on third parties and reputational risks.²

This approach, based on principles and on ordinary supervision, contrasts with the European regulation on artificial intelligence, which entered into force in 2024. That regulation classifies systems by level of risk and imposes detailed obligations on high-risk systems, which include certain financial uses such as creditworthiness assessment.³ At the beginning of 2025 the Federal Council chose not to adopt a general statute on artificial intelligence, to ratify the Council of Europe convention on artificial intelligence and to proceed by sectoral adjustments. This approach is consistent with the one adopted for crypto and digital assets.⁴ For Geneva’s institutions the framework is therefore that of FINMA, supplemented by the data protection act, revised and in force since 1 September 2023, and by the European rules where they serve clients in the Union.⁵

Notes

1. FINMA, supervisory communication 08/2024 of 18 December 2024, “Governance and risk management when using artificial intelligence”, finma.ch. ↩
2. FINMA, supervisory communication 08/2024, cited above. ↩
3. Regulation (EU) 2024/1689 of 13 June 2024 on artificial intelligence, published in the Official Journal on 12 July 2024, in force since 1 August 2024, applicable in stages: prohibited practices from 2 February 2025, general-purpose models from 2 August 2025, high-risk systems from 2 August 2026 and, for regulated products, from 2 August 2027; Direction générale des Entreprises (France), “Le Règlement européen sur l’intelligence artificielle”, entreprises.gouv.fr (accessed September 2026). ↩
4. Federal Council, “Réglementation de l’IA : le Conseil fédéral veut ratifier la Convention du Conseil de l’Europe” (AI regulation: the Federal Council wishes to ratify the Council of Europe Convention), press release of 12 February 2025, admin.ch: ratification of the convention and sectoral adjustments to national law. ↩
5. Federal Act of 25 September 2020 on Data Protection (FADP; SR 235.1), in force with its ordinances since 1 September 2023; Federal Office of Justice, “Nouveau droit de la protection des données”, bj.admin.ch (accessed September 2026). ↩

For general information only. This article does not constitute legal or tax advice, does not replace advice on an individual case and creates no lawyer-client relationship. It reflects the legal position as of September 2025.