MAS FEAT Principles for AI and Data Analytics

What it is

The Principles to Promote Fairness, Ethics, Accountability and Transparency (FEAT) in the Use of Artificial Intelligence and Data Analytics in Singapore’s Financial Sector were published by the Monetary Authority of Singapore on 12 November 2018. They give financial firms a set of foundational principles for decisions made or assisted by artificial intelligence and data analytics, which the document calls AIDA.

The principles are not prescriptive and are not intended to replace a firm’s existing internal governance. MAS described the PDPC’s Model AI Governance Framework as complementary to them.

Who it applies to

All firms using AIDA to provide financial products and services in Singapore, for both complex and straightforward techniques. Firms can calibrate the actions the principles call for to the materiality of their AIDA-driven decisions, considering factors such as how far AIDA is used in the decision, the model’s complexity, the degree of automation, the severity and probability of impact on stakeholders, and the options for recourse. MAS encourages firms to read the principles as a whole, because they reinforce each other.

What it requires

There are 14 principles across the four headings.

Fairness. Justifiability and accuracy and bias.

1. Individuals or groups are not systematically disadvantaged through AIDA-driven decisions unless those decisions can be justified.
2. Use of personal attributes as input factors is justified.
3. Data and models are regularly reviewed and validated for accuracy and relevance, and to minimise unintentional bias.
4. AIDA-driven decisions are regularly reviewed so models behave as designed and intended.

Ethics. 5. Use of AIDA is aligned with the firm’s ethical standards, values and codes of conduct. 6. AIDA-driven decisions are held to at least the same ethical standards as human-driven decisions.

Accountability. Internal and external. 7. Use of AIDA in decision-making is approved by an appropriate internal authority. 8. Firms are accountable for both internally developed and externally sourced AIDA models. 9. Firms proactively raise management and board awareness of their use of AIDA. 10. Data subjects have channels to enquire about, appeal and request reviews of AIDA-driven decisions that affect them. 11. Verified and relevant supplementary data from data subjects is taken into account in a review.

Transparency. 12. Use of AIDA is proactively disclosed to data subjects as part of general communication. 13. Data subjects receive, on request, clear explanations of what data is used to make AIDA-driven decisions about them and how it affects the decision. 14. Data subjects receive, on request, clear explanations of the consequences those decisions may have for them.

The Veritas methodology

MAS then worked with the industry through Veritas, a multi-phase initiative, to make the principles assessable. Phase 1 concluded on 6 January 2021 with fairness metrics and assessment methods for credit risk scoring and customer marketing. Phase 2 concluded on 4 February 2022 with the full FEAT assessment methodology and the first Veritas Toolkit. Phase 3 concluded on 26 June 2023 with Toolkit version 2.0 and pilots integrating the methodology into the governance frameworks of seven financial institutions. The toolkit is open source.

Key dates

• 12 November 2018. FEAT Principles published.
• 6 January 2021. Veritas phase 1 concludes.
• 4 February 2022. Veritas phase 2 concludes.
• 26 June 2023. Veritas phase 3 concludes.
• 7 October 2026. The MAS AI Risk Management Guidelines confirm FEAT continues to apply.

How it relates to other frameworks

FEAT is the principles layer for AI in Singapore’s financial sector. The MAS Guidelines on AI Risk Management, issued in October 2026 and in effect from October 2027, now carry the supervisory expectations for how institutions govern AI, and they refer back to FEAT for what fair outcomes mean. An institution that has already assessed its systems against FEAT can use that work as input to its fairness controls under the Guidelines.

The external accountability and transparency principles cover similar ground to data protection rules on automated decisions elsewhere, such as GDPR Article 22’s safeguards and Australia’s Privacy Act ADM transparency obligations.

A note on this page

This is Aicura’s reading of the FEAT Principles, written to help you understand them. It is not legal advice and it is not the document itself. Read the primary source from MAS, and take advice on how it bears on your institution.