How does OCBC govern AI responsibly?
OCBC's stated standard is MAS's FEAT principles of fairness, ethics, accountability and transparency. Bank of Singapore said in October 2025 that all in-house AI tools are built in line with FEAT and regularly reviewed for accuracy and to prevent bias across gender, nationality and other groups; the then group COO described the same reviews in 2023. The bank builds most tools itself, which Group Data Office head Donald MacDonald argued in 2024 makes decisions easier to explain to regulators and customers, and in 2024 it was keeping generative AI away from customers until its guardrails were robust enough. Its agentic tools keep people on decisions: relationship managers review every Source of Wealth Assistant draft, and HELIOS adds an independent checker agent while bankers and review teams keep 'ultimate accountability'. Staff and client data for OCBC GPT and SOWA stay on private clouds. OCBC sat on the MAS consortium behind the Veritas toolkit (2023), the MindForge toolkit (March 2026) and the SAFR agent-safeguards paper (July 2026), and its head of operations and technology said in April 2026 that all frontier AI tools undergo 'rigorous assessment and validation before deployment'. MAS's final AI Risk Management Guidelines, issued October 7, 2026, apply from October 2027.[1][2][3][4][5][6][7][8][9][10]
Data as of Oct 16, 2026 · OCBC profile updated Oct 13, 2026 · 21 sources cited on this page · 140 banks profiled on this site · Markdown
What controls does OCBC apply to its AI systems?
Controls the bank has described, by system and date.
| Control | Where applied | Date |
|---|---|---|
| FEAT model reviews: accuracy tracked over time; tests that models are not biased against any group, gender or nationality | All AI models | November 2023; restated October 2025 and March 2026[2][1][11] |
| Private hosting: information processed stays on a private cloud dedicated to the bank | OCBC GPT; Source of Wealth Assistant | October 2023; October 2025[5][1] |
| No customer-facing Gen AI until guardrails could contain hallucination and adversarial-prompting risks | Generative AI generally | November 2024[3] |
| Human review of every AI draft before anti-money-laundering assessment | Source of Wealth Assistant | October 10, 2025[1] |
| Independent checker agent; humans verify what it cannot; bankers keep accountability for decisions | HELIOS | July 29, 2026[12][4] |
| Avatars not licensed to advise; likeness removed if the employee leaves; consent can be withdrawn | OCBC WoW | September 2026[13] |
| Deterministic guardrails so avatars draw only on OCBC's proprietary research | OCBC WoW | July 2026[14] |
| 'Clear checkpoints, validation steps and escalation paths' for agentic workflows | Agentic AI generally | March 2026[11] |
How has OCBC worked with MAS and other Singapore authorities on AI?
Industry programmes and supervisory contact on the public record.
| Initiative | OCBC's role | Date |
|---|---|---|
| Veritas toolkit 2.0 (MAS) | One of 31 consortium members behind the open-source FEAT assessment toolkit | June 27, 2023[6] |
| Generative AI evaluation sandbox (IMDA, AI Verify Foundation) | One of 15 initial participants, as an application developer | November 1, 2023[15] |
| Pre-launch walkthrough | A Bank of Singapore data-science lead took MAS through the safeguards for its agentic drafting tool before launch | Reported January 8, 2026[16] |
| MindForge AI risk-management toolkit (MAS) | One of 24 consortium members | March 20, 2026[7] |
| SAFR white paper (MAS) | Case study on runtime safeguards for AI agents | July 3, 2026[8] |
| AI Risk Management Guidelines (MAS) | Applies to OCBC like all Singapore financial institutions; sections 3 and 4 from October 7, 2027, sections 5 and 6 from October 7, 2028 | October 7, 2026[10] |
What does OCBC not publish about its AI governance?
OCBC has not published a stand-alone AI principles document of its own; it cites MAS's FEAT principles. It discloses that reviews happen but not their results: no count of models reviewed, no bias-test outcomes and no named AI ethics committee. The 2025 annual report places responsible-AI governance in its risk chapter, under the Group Chief Risk Officer's remit.[1][17]
Third-party model risk is addressed in general terms. After Anthropic's Claude Mythos Preview raised fears of faster vulnerability discovery, Praveen Raina said in April 2026 that all such solutions undergo rigorous assessment and validation before deployment, and the bank has not named the large language models behind OCBC WoW.[9][14]
What does the OCBC profile record on this topic?
The rows below are the entries of the OCBC Bank AI strategy profile that bear on this page, as of Oct 13, 2026; the profile holds the full record of 24 dated moves, 14 use cases and 32 sources.
| Use case | What the record shows | Status | Source |
|---|---|---|---|
| AI copilot for corporate relationship managers | Generates ESG, financial and industry analysis from lengthy unstructured documents during credit assessment, cutting preparation time by up to 30%. | In production | [17] |
| FEAT-based AI governance and MAS toolkit participation | Regular model reviews for accuracy and bias, tested against fairness, ethics, accountability and transparency (FEAT) principles; OCBC took part in MAS's MindForge toolkit and the SAFR agent-safeguards paper. | In production | [18][7][8][17] |
| Date | Move | Detail | Source |
|---|---|---|---|
| Oct 7, 2026 | MAS issues final AI Risk Management Guidelines | The guidelines apply to all financial institutions, including OCBC, and take effect on 7 October 2027 for sections 3 and 4 and 7 October 2028 for sections 5 and 6. They cover board oversight, lifecycle controls and accountability for third-party AI; MAS plans to consult on agentic AI guidance in 2027. | [10] |
| Jul 3, 2026 | MAS publishes SAFR white paper with an OCBC case study | Safeguards for Agentic Finance at Runtime proposes governance checkpoints that verify and record an AI agent's proposed actions before execution; OCBC is among the firms whose case studies it contains. | [8] |
| Apr 30, 2026 | Praveen Raina says OCBC assesses AI tools before deployment | Reuters reported Asian banks tightening checks on AI tools after Anthropic's Claude Mythos Preview raised fears that attackers could find weaknesses faster. OCBC's head of group operations and technology said in a statement that all such solutions undergo rigorous assessment and validation before deployment. | [9] |
| Mar 20, 2026 | MAS AI risk-management toolkit includes OCBC as a participant | OCBC is among 24 consortium members, with DBS, UOB, Standard Chartered and others, behind the MindForge toolkit covering traditional, generative and agentic AI. | [7] |
| Nov 13, 2025 | MAS publishes AI risk-management guidelines for consultation | Proposed guidelines for all financial institutions expect board and senior-management oversight, AI inventories, risk-materiality assessments and proportionate controls on data, fairness and human oversight. | [19] |
| Oct 10, 2025 | Bank of Singapore announces agentic Source of Wealth Assistant | The tool drafts know-your-customer source-of-wealth reports from client documents, cutting preparation from 10 days to an hour; relationship managers review the draft before internal assessment. The 2025 annual report says it launched in June 2025 in collaboration with OCBC's group data office. | [20][17] |
| Feb 26, 2025 | Helen Wong tells analysts AI will create more jobs | On the full-year 2024 results call, Wong said OCBC had built its own internal AI system for coding and ranking work that helped coders work 20% faster, and gave hiring more staff to help engineers add features to ATMs as an example of AI creating jobs. OCBC was also building engineering hubs in Shanghai and Shenzhen and expanding its Indonesia hub. | [21] |
What AI principles does OCBC follow?
OCBC says its AI tools are built in line with the FEAT principles of fairness, ethics, accountability and transparency set by the Monetary Authority of Singapore. Bank of Singapore said in October 2025 that all in-house AI tools undergo regular reviews to ensure accuracy and prevent bias across gender, nationality and other groups. OCBC has not published a separate AI principles document of its own.
Does OCBC keep humans in the loop on agentic AI?
Yes, on the tools it has described. Relationship managers review and refine every Source of Wealth Assistant draft before it goes to compliance review, and for HELIOS an independent agent checks the others while relationship managers and review teams keep ultimate accountability for judgment and decisions. Strategy chief Melvyn Low has called for clear checkpoints, validation steps and escalation paths in agentic workflows.
How do MAS's AI guidelines affect OCBC?
MAS issued final AI Risk Management Guidelines on October 7, 2026 for all financial institutions in Singapore, including OCBC. They cover board oversight, lifecycle controls and accountability for third-party AI, with sections 3 and 4 applying from October 7, 2027 and sections 5 and 6 from October 7, 2028. MAS plans to consult on agentic AI guidance in 2027.
Has OCBC worked with MAS on AI safety?
Yes. OCBC was one of 31 members of the MAS-led consortium behind the Veritas responsible-AI toolkit in 2023, one of 24 behind the MindForge AI risk-management toolkit in March 2026, and supplied a case study for the SAFR white paper on safeguarding AI agents in July 2026. A Bank of Singapore team walked MAS through the safeguards on its agentic drafting tool before launch.
- Bank of Singapore deploys agentic AI tool to automate writing of source of wealth reports
- Generative AI, the buddy that OCBC taps to deliver better experiences for customers and employees (Future of Finance 2023)
- Transforming the banking experience with generative AI (Future of Finance 2024)
- OCBC harnesses agentic AI to sharpen and quicken onboarding of wealthy customers
- OCBC to deploy generative AI bot for all 30,000 staff globally
- AI ethics toolkit updated to include more assessment components
- MAS launches AI risk toolkit for financial institutions with case studies from DBS, peers
- MAS partners finance industry players to develop safeguards for AI agents
- Banks in Asia brace for complex cyber threats from frontier AI
- MAS rolls out AI risk management guidelines for Singapore's financial sector
- Inside local bank's AI engine: From smarter security to personalised banking, at scale (OCBC branded content)
- OCBC to use agentic AI to cut wealth onboarding time to 15 business days
- Meet the humans behind 'Wendy and Wayne', OCBC's wealth avatars
- OCBC rolls out AI avatars for wealth management (undated page, retrieved 2026-10-05)
- Global players look to create baseline to evaluate generative AI applications
- Inside Singapore's AI bootcamp to retrain 35,000 bankers
- OCBC Annual Report 2025
- Annual Report 2024 | Creating Value Through AI, Transforming for the Future (undated page, retrieved 2026-10-05)
- MAS to trial issuance of tokenised bills, releases guidelines for AI in financial sector
- Bank of Singapore's new AI tool does in one hour what used to take 10 days
- OCBC CEO Helen Wong thinks AI will create more banking jobs
When OCBC moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am ET · free
plus every tracker, bank and agent page update, the morning after · leave any morning