Suppose you ask your bank's in-app assistant whether a transfer fee can be waived, and it says yes. You act on it and the fee is charged. Who carries the consequence? The institution. In its 7 October 2026 media release, MAS stated that financial institutions remain accountable for AI used in the services they deliver, including AI developed, operated or provided by third parties. A vendor may have built the chatbot, but the bank put its name to the service. [1]
Two dates shape what happens next. The Guidelines on Artificial Intelligence Risk Management take effect on 7 October 2027, with Sections 5 and 6 to be met by 7 October 2028, so they set what institutions must have in place rather than what you can use this month. What already applies is MAS's Guidelines on Fair Dealing: customers should receive clear, relevant and timely information that accurately represents the products and services offered and delivered, and complaints should be handled independently, effectively and promptly. The Straits Times reported on the accountability point that same week. [1] [2] [3]
Whose answer was it?
That sentence does more work than it looks. MAS pairs it with duties that sit with the bank rather than the customer: obtain sufficient assurance from third-party providers, assess whether third-party AI suits its intended use, apply compensating controls where assurance gaps arise, and consider limiting, suspending or replacing the service if risks cannot be brought within the bank's risk appetite. Control effort then follows risk exposure, with basic policies and procedures allowed where poor performance or unavailability of an AI tool is unlikely to have a material impact on the institution, its customers or other stakeholders. [1]
Continuity matters more than novelty here. The Guidelines give the example that an institution should continue to adhere to MAS's Guidelines on Fair Dealing even if it leverages AI to deliver products and services. The idea is not alien to MAS either: its 2018 digital advisory guidelines already expect digital advisers to disclose in writing the assumptions, limitations and risks of their algorithms, and the circumstances in which they may override them or halt the service. That guideline was written for advisers giving investment advice, so read it as comparison, not as a rule for a general banking chatbot. [4] [5]
What the bank is expected to do about it
The machinery starts with naming the tool. MAS expects institutions to identify AI use across business and functional areas, and where AI is embedded in, or used to deliver, third-party services, to minimally cover such AI from material third-party service providers. Institutions should then maintain an accurate inventory of AI use cases, systems or models, to the extent possible or practicable. Reputational risk comes with a customer-facing example in the Guidelines: chatbots could provide incorrect information. [4]
Named use cases are then assessed on materiality. The assessment minimally covers impact — the potential consequences of a failure, malfunction or poor performance for the institution and for customers or other stakeholders, including consumer protection — and reliance, meaning how far the AI is relied on for the output, the autonomy granted and the degree of human oversight. Senior management is expected to keep an internal escalation process for material risks, including AI incidents, which the Guidelines define as failures, malfunctions or poor performance of the underlying system or model. Where a use case is assessed as high risk, contingency plans should set out fallbacks such as alternative systems or manual processes. [4]
The complaint route, in order
Start where MAS starts. In a 2022 letter responding to a Straits Times reader, MAS said that consumers with a complaint against any financial institution should first contact the institution directly, and that it expects all institutions to have in place a process to resolve consumer complaints promptly and effectively. If you are not satisfied with the response, you may approach the Financial Industry Disputes Resolution Centre, an independent and impartial body providing low-cost dispute resolution services for consumers and their financial institutions. [6]
The FIDReC conditions in the terms of reference published on MAS's site are specific: the institution must have failed to resolve the complaint to your satisfaction within eight weeks of receiving it; an application can be made at the earliest after at least eight weeks from the date the matter was referred to the institution, and at the latest six months after its final reply has been received by you. A reply counts as final only if the letter or document expressly says it is the institution's final reply. That published copy shows no date, so confirm the current timings with FIDReC before relying on them. [7]
Escalating, and keeping the record
Separately from a dispute, MAS's 2022 letter says consumers should file a report with MAS if they suspect that a financial institution or its representative has committed a regulatory breach or misconduct, and that MAS will investigate. MAS does not resolve individual disputes. Two boundaries are worth stating plainly. The AI Guidelines are supervisory expectations on institutions, not a new consumer remedy, and nothing here assesses an individual case. The erroneous-transaction provisions of the E-Payments User Protection Guidelines apply to payment accounts where the institution is the recipient's financial institution, so they do not establish that a wrong chatbot answer counts as an erroneous transaction. [6] [8]
As an optional step, you might keep the date and time, the exact wording the assistant gave, a screenshot of the exchange and what you did next, then ask the bank to treat it as a formal complaint and to give you a reference and a written response. No published figure shows how often banking assistants answer wrongly or how such complaints end, so treat that as preparation, not a prediction. Either way the answer to your question does not change: the institution is accountable for the answer its AI gives, including when a vendor built the tool; the complaint starts with the bank; and the clocks that matter for FIDReC run from there. [2] [6]
Read next
Sources
- MAS Sets Out Supervisory Expectations on Responsible AI Adoption by Financial Institutions
- Guidelines on Fair Dealing
- MAS guidelines on third-party AI use in financial institutions | The Straits Times
- [PDF] Guidelines on Artificial Intelligence Risk Management
- SECURITIES AND FUTURES ACT (CAP. 289) FINANCIAL ADVISERS ACT (CAP. 110)
- Response to Letter "Proper channel needed for complaints against insurers" – The Straits Times 14 October 2022
- TERMS OF REFERENCE
- Monetary Authority of Singapore
BUTLER Magazine Editorial · AI-assisted research and writing, reviewed by our automated editorial team. Sources checked 2026-10-09. Featured image: AI-generated editorial illustration.