Comparison with Geneva, Zurich and the European Union

Read the PDF edition (2 pages)

Switzerland and Singapore share a preference for principles and ordinary supervision over a dedicated statute. The Swiss Financial Market Supervisory Authority (FINMA) set out its expectations in Guidance 08/2024 of 18 December 2024. It asked supervised institutions to identify, assess, manage and monitor the risks of artificial intelligence and found that most were “still in the early stages of development”, with governance structures still being established.¹ The two regulators observed the same stage of adoption at almost the same time. MAS then went further, converting its observations into detailed guidelines with transition periods, whereas FINMA has so far relied on its guidance and on ordinary supervision.

The European Union chose legislation. Its Artificial Intelligence Act entered into force on 1 August 2024 and applies in stages. Prohibited practices became unlawful from 2 February 2025 and the rules on general-purpose models applied from 2 August 2025. The obligations for high-risk systems listed in Annex III, which include certain financial uses, were postponed by the digital omnibus regulation of 2026 to 2 December 2027.² Geneva’s banks feel the effect of the European regime when they serve clients in the Union. Singapore’s institutions feel it mainly through European parent companies and clients.

The difference in method reflects a difference in the role of the regulator. In Geneva, the Geneva volume describes a cautious adopter that waits for proven technology, with a financial centre of about 38,000 jobs.³ In Singapore, MAS acts as convenor of industry consortia, funder of pilots and author of guidelines, a combination that the Swiss authorities do not attempt. Comparable principles exist elsewhere in Asia and in Europe. The Hong Kong Monetary Authority issued guiding principles on big data analytics and artificial intelligence in 2019, as did the Dutch central bank with its SAFEST principles. Hong Kong’s wider position as a competitor is examined in Chapter K.III.⁴

The comparison yields a lesson for both centres. Geneva’s prudence avoids the costly failures of pioneers, while it risks a growing distance from centres that learn by doing. Singapore’s activism produces common methods and a trained workforce, at the price of a regulator closely involved in the industry it supervises. For a client choosing between the two centres, the more relevant question is whether the institution concerned can explain how its models work and who answers for them. On that question, the Singapore Guidelines give the client a clearer standard to invoke.

Notes

1. FINMA, “FINMA guidance on governance and risk management when using artificial intelligence”, FINMA Guidance 08/2024, 18 December 2024 (finma.ch): risks include model risks (robustness, correctness, explainability, bias), data risks, IT and cyber risks, third-party dependence, legal and reputational risks; most institutions are “still in the early stages of development”. ↩
2. Wikipedia, “Artificial Intelligence Act” (accessed November 2026): published in the Official Journal on 12 July 2024, in force on 1 August 2024; prohibited practices from 2 February 2025; general-purpose AI rules from 2 August 2025; Regulation (EU) 2026/1744 (digital omnibus), in force on 27 July 2026, postponing Annex III high-risk obligations to 2 December 2027 and Annex I obligations to 2 August 2028. ↩
3. Ressos, Genève, place financière internationale, Part K, chapter IV, section on employment (citing Léman Bleu, 16 October 2025): nearly 38,000 jobs in the Geneva financial sector. ↩
4. MAS, Information Paper, December 2024, footnote 21: the HKMA issued guiding principles for big data analytics and AI in 2019 covering governance and accountability, fairness, transparency and disclosure, and data privacy; De Nederlandsche Bank issued the SAFEST principles in 2019. ↩

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 October 2026.