AI regulation in wealth management sounds like a settled subject; it is anything but. Ask a supervisor whether a wealth manager may use artificial intelligence and you will get a version of the same answer everywhere: yes — and it is your problem when it goes wrong. The interesting question in 2026 is therefore not whether AI is allowed. It is who is accountable when the model is wrong. On that point, four financial regulators that matter to Swiss wealth management have arrived at instructively different ways of saying the same thing.
This article compares how four supervisors approach AI regulation in wealth management: FINMA in Switzerland, the FCA in the United Kingdom, MAS in Singapore and the FSRA in Abu Dhabi Global Market, with the EU AI Act as the binding counter-example. Every claim links to the regulator’s own publication, with dates, because this field moves monthly. Everything below was checked against the primary sources on 17 August 2026.
AI regulation in wealth management at a glance
| Regulator | Approach | Binding AI-specific rules | Testing environment | Centre of gravity | Key document (date) |
|---|---|---|---|---|---|
| FINMA (Switzerland) | Principles-based, technology-neutral supervision | No — supervisory guidance within existing law | None dedicated to AI | Governance and model risk | Guidance 08/2024 (18 December 2024) |
| FCA (United Kingdom) | Outcomes-based; existing framework applied to AI | No — explicitly no new AI rulebook so far | AI Lab: Supercharged Sandbox, AI Live Testing | Consumer outcomes and senior-manager accountability | AI Update (April 2024); AI Lab (from October 2024) |
| MAS (Singapore) | Principles first, now moving to formal guidelines | Not yet — guidelines under consultation | Veritas; pathfin.ai; joint work with the FCA | Fairness and lifecycle risk management | FEAT Principles (November 2018); AI risk management consultation (13 November 2025) |
| FSRA, ADGM (Abu Dhabi) | Technology-neutral rulebook plus joint UAE guidance | No | ADGM innovation agenda | Enabling adoption within existing rules | Joint Guidelines on Enabling Technologies (15 November 2021) |
| EU (contrast) | Statutory, risk-based product regulation | Yes — staged entry into application | Member-state regulatory sandboxes | Risk classification and conformity | AI Act (in force 1 August 2024) |
Switzerland: how FINMA regulates AI in wealth management
FINMA has not written an AI regulation for wealth management or anyone else, and has said it does not need one. Its Guidance 08/2024, published on 18 December 2024, sets out supervisory expectations for institutions using AI: know where it is used, classify the risks — robustness, correctness, explainability, bias, data quality and security, growing third-party dependence — assign responsibilities, test before deployment and monitor afterwards. A guidance note is not a new law; it tells supervised firms how FINMA reads existing duties, in the same family as its requirements on operational risk and outsourcing. The wider Swiss rulebook for wealth managers is mapped in our overview of regulations in Swiss wealth management.
What this means in practice: a licensed portfolio manager does not need FINMA’s permission to use AI. It needs to be able to show, on request, which tools are in use, for which purposes, with which risks, under whose responsibility. An undocumented chatbot habit in the client team is exactly the situation the guidance is aimed at — accountability that evaporated into a tool nobody officially introduced.
United Kingdom: the FCA bets on its existing framework
The FCA has said, repeatedly and in writing, that it is not planning an AI rulebook for now. Its April 2024 AI Update set the line, and its current approach page maintains it: existing, outcomes-based rules — the Consumer Duty, the Senior Managers and Certification Regime — already reach AI-driven activity. What the FCA has built instead is infrastructure for learning: the AI Lab, launched in October 2024, with a Supercharged Sandbox run with NVIDIA whose first cohort completed and presented in January 2026, an AI Live Testing programme whose second cohort was selected in April 2026, and a public call for examples of good and poor practice that closed in June 2026, with findings promised later in the year. A separate review, the Mills Review, is examining the long-term effect of AI on retail financial services.
What this means in practice: in UK wealth management, AI regulation works through personal accountability. A named senior manager answers for outcomes produced with AI, and the model decided is not a defence under the SM&CR. For a Swiss firm with UK clients, that framing travels well: it is close to how FINMA thinks, expressed through individual accountability rather than institutional governance.
Singapore: MAS moves from principles to guidelines
MAS started earliest and is now moving furthest. Its FEAT Principles — fairness, ethics, accountability, transparency — date from November 2018, and the Veritas programme spent years turning them into assessment methodology. On 13 November 2025 MAS published a consultation paper on Guidelines on Artificial Intelligence Risk Management; the consultation closed on 31 January 2026, and a twelve-month transition period is planned once the guidelines are finalised. At the time of writing the final text is still pending. The draft expects central AI inventories, risk assessment along impact, complexity and reliance, controls across the whole lifecycle, and explicit attention to generative AI and AI agents — with the board and senior management carrying ultimate responsibility.
What this means in practice: for wealth management, Singapore is producing the most concrete AI regulation of the four. A firm serving Singapore-connected clients should start its AI inventory now, because retrofitting one under a twelve-month transition clock is unpleasant — a lesson Swiss firms already know from the evolution of KYC.
Abu Dhabi: the FSRA enables first, regulates through existing rules
The FSRA in Abu Dhabi Global Market has issued no AI-specific rulebook. The anchor document is older and broader: the Guidelines for Financial Institutions Adopting Enabling Technologies of 15 November 2021, issued jointly by the UAE central bank, the securities regulator, Dubai’s DFSA and the FSRA. They cover AI and big data alongside APIs, biometrics, cloud and distributed ledgers — cross-sectoral principles and best practice, not binding rules. Around this sits national industrial policy, the UAE National AI Strategy 2031, and an ADGM innovation agenda that includes a collaboration with the Mohamed bin Zayed University of Artificial Intelligence announced in 2025. One distinction worth keeping sharp: ADGM in Abu Dhabi and the DIFC in Dubai are separate jurisdictions with separate regulators; the FSRA is not the DFSA.
What this means in practice: the posture is promotional — attract AI-driven business, supervise it through technology-neutral rules. For wealth management in ADGM, AI regulation in practice means the firm’s own governance plus the joint guidelines’ expectations on risk management.
The EU AI Act: the binding counter-example
The EU chose the opposite instrument: a statute. The AI Act entered into force on 1 August 2024 and applies in stages — prohibited practices and AI-literacy duties from February 2025, obligations for general-purpose models from August 2025. The original date for most high-risk obligations, 2 August 2026, did not survive contact with reality: under the digital omnibus package agreed in 2026, key high-risk deadlines are being deferred towards late 2027, subject to formal adoption. For wealth management the classification matters more than the calendar: creditworthiness assessment sits in the high-risk annex, while portfolio management and investment advice as such generally do not — which leaves transparency duties, prohibited practices and the general-purpose model regime as the relevant layers.
For a Swiss firm serving EU clients, the Act can reach across the border when AI output is used in the Union. It joins a long list of cross-border rules that a Swiss compliance framework has to absorb without an EU passport to show for it.
The thread that connects all four
Strip away the acronyms and the four non-EU regulators are saying one thing in four dialects: no supervisor accepts the model decided as an answer. FINMA locates responsibility in governance and the board. The FCA pins it on a named senior manager. MAS writes board and senior management into the first chapter of its draft guidelines. The FSRA leaves existing accountability rules to do the work. The EU adds product-style conformity duties on top of the same idea. Method differs — guidance, outcomes, guidelines, statute — the destination does not: human, documented, institutional accountability. That, and little else, is the current content of AI regulation in wealth management. The cheapest compliance strategy available is to behave as if your regulator had already adopted the strictest version.
AI regulation in practice: what a Swiss wealth management firm should document
- An inventory of every AI tool in use, including the unofficial ones.
- Purpose and risk classification per use case — client-facing uses first.
- A named owner for each tool and use case.
- Data flows: what goes in, where it is processed and stored, under which data processing agreement.
- Vendor due diligence and an exit scenario, in line with outsourcing expectations.
- Defined human review points for anything that reaches a client, a regulator or a portfolio.
- A testing and monitoring log, however modest.
- Staff instructions and training, in writing.
Where the tools themselves are the question, our comparison of AI tools for wealth managers and our overview of the Swiss wealthtech landscape are the companion pieces.
FAQ
Does a Swiss portfolio manager need FINMA approval to use AI?
No. There is no authorisation requirement for the technology itself. FINMA’s Guidance 08/2024 instead expects documented governance: inventory, risk assessment, responsibilities, testing and monitoring, proportionate to the firm.
Which AI regulation applies to a wealth management firm with clients abroad?
Primarily the Swiss one — but local conduct rules can attach to the client relationship, and the direction of travel abroad shapes what counts as sound practice at home. Cross-border service has never been a purely domestic question.
Is the EU AI Act relevant to a firm with no EU office?
Potentially, yes. Its reach is extraterritorial where AI output is used in the EU. Whether that bites depends on the use case and its risk classification, and the deadlines have been moving — verify the current state before building anything on it.
Will the four approaches converge?
On substance they already have: accountability, inventories, human oversight. On instruments, unlikely — a guidance note, an outcomes regime, formal guidelines and a statute age differently. Expect the vocabulary to converge before the paperwork does.
All regulatory documents cited were checked against the regulators’ own websites on 17 August 2026. This field moves quickly; verify the current status of every document before relying on it. This article reflects personal views and is not legal advice.
