Short answer: A responsible AI checklist for a small business should cover seven things: know where AI is used, name an owner, protect personal and confidential data, check outputs for errors and bias, keep a human in the loop where decisions affect people, be open with customers and staff about AI use, and keep records. These map to themes in Singapore's voluntary guidance, such as the IMDA Model AI Governance Framework and the AI Verify principles.
Why does an SME need a checklist?
Large organisations have risk teams. Small ones usually do not, yet staff may already use AI for emails, quotes, marketing and customer replies. A short checklist turns principles into routine habits and shows you took reasonable care.
What should the checklist include?
1. Do we know where AI is used? Make a list, including tools that staff signed up for themselves. You cannot manage what you have not listed.
2. Who is accountable? Name one person. Accountability is a recurring theme in governance guidance: the IMDA Model AI Governance Framework has a section on internal governance structures and measures, and IMDA's framework for agentic AI says human responsibility continues to apply even when an AI agent acts on its own.
3. Is personal and confidential data protected? Check what data goes into each tool and whether it is covered by PDPA obligations. The PDPC's Advisory Guidelines on the Use of Personal Data in AI Recommendation and Decision Systems (issued 1 March 2024) explain how the PDPA applies when personal data is used in AI systems, including consent, notification and accountability. Check vendor terms on storage and reuse of your data.
4. Are outputs checked? AI can produce confident but wrong answers. Decide which outputs need checking before use: legal text, prices, medical or financial information, and anything sent to a customer.
5. Is there a human in the loop? The more an automated decision affects a person's rights or opportunities, the more human review it deserves. The IMDA Model AI Governance Framework has a section on determining the level of human involvement in AI-augmented decision-making.
6. Are we transparent? Tell people when they are dealing with an AI system if it matters to them, and say how to reach a human.
7. Do we keep records? Note what the tool is for, what you tested, what you decided and when you last reviewed it.
How do we test fairness without a data science team?
For small firms, simple checks go a long way. Run sample inputs that differ only by an attribute such as name, gender or age and see whether outputs change in ways that cannot be justified. For anything more serious, such as hiring or credit, seek specialist advice. Financial institutions should note that MAS issued Guidelines on Artificial Intelligence Risk Management on 7 October 2026, which build on its FEAT principles of fairness, ethics, accountability and transparency.
How often should we review?
At least once a year, and whenever you change a tool, a vendor, or a use case. Also review after any incident or complaint.
What does the checklist not cover?
It is not a compliance certificate. Regulated sectors have additional expectations, and a lawyer should advise on legal duties. Voluntary frameworks are evolving, so recheck the sources below.
One-page version
- List AI tools: done / not done
- Owner named: done / not done
- Data rules written: done / not done
- Output checking rules: done / not done
- Human review points set: done / not done
- Transparency wording agreed: done / not done
- Records kept and review date set: done / not done
Keep learning
Join the free monthly newsletter at asia-ai.org. For members, see asia-ai.org/join for current benefits.
This article is general information, not legal advice.