Fairness, Transparency, and Oversight

Sections 3, 4, and 7 of the CBUAE Guidance Note give recommendations on fairness and non-discrimination, transparency and explainability, and human oversight. The Central Bank of the UAE (CBUAE) issued the Guidance Note on 11 February 2026. The Guidance Note gives advisory principles, and it supplements applicable laws, regulations, and directives without replacing them: see The CBUAE Guidance Note. The Guidance Note addresses UAE licensed financial institutions (LFIs), "including insurance providers", and consumer-related AI/ML uses (Scope and AI uses).

Fairness and non-discrimination

Section 3 of the CBUAE Guidance Note gives LFIs these recommendations on fairness and non-discrimination:

• Outcomes, Section 3(a). LFIs "are expected to ensure that AI and ML systems do not result in discriminatory or manipulative outcomes against individuals or groups." Section 3(a) adds: "No AI system or application should be deployed or used if it is discriminatory or manipulative or develops as such post-deployment."
• Training data, Section 3(b). The data that LFIs use to train AI and ML models "should be sufficiently accurate, relevant and representative of the customer populations to which the models will be applied."
• Bias tests, Section 3(c). Deployed AI "should be subject to periodic testing". The tests occur "once a year or each time a model is upgraded, materially changed or a new one is introduced". The purpose of the tests is "to identify and remediate undue/unintended embedded biases or discriminatory outcomes."
• Ethical standards, Section 3(c). The deployment of AI "should reflect the institution's ethical standards and code of conduct." Section 3(c) adds: "Decisions made or supported by AI should be consistent with the duty to act honestly, fairly and in the best interests of consumers."

Section 6(c) gives LFIs recommendations on testing, awareness, and output bias for automatic updates (Model validation and automatic updates).

Transparency and explainability

Section 4 of the CBUAE Guidance Note gives LFIs recommendations on transparency and explainability. Section 2(f) adds a tool for explanation.

Transparency about AI use, Section 4(a). Section 4(a) recommends that LFIs are transparent with customers and "relevant stakeholders" about the use of AI. The recommendation is about the use of AI in general. Section 4(a) puts emphasis on high-impact decisions. It also covers customers and stakeholders "if they are communicating or interacting with an AI application".

A high-impact decision is a determination by an LFI with AI that "materially affects a customer's access to financial products or services": see Definitions.

Explanation of AI decisions, Section 4(a). Section 4(a) also gives this recommendation: "LFIs should be clear as to how AI systems operate and make decisions and be able to disclose the same." The Guidance Note does not name an internal explanation record as a replacement for this disclosure.

Language of disclosures, Section 4(b). Section 4(b) gives this recommendation: "Understandable plain language and accurate disclosures should be made in both Arabic and English, with telephone support available in all major languages of the UAE." Section 4(b) adds: "LFIs should consider the use of appropriate measures to check understandability." The Guidance Note recommends that LFIs "consider" these measures. It does not recommend a check of the understandability of each disclosure.

Opt-out rights, Section 4(c). Section 4(c) gives this recommendation:

LFIs should consider the provision of opt-out rights with respect to AI for customers, particularly for High-impact decisions, taking into account the potential risks or impact to the customer, fairness and feasibility.

Section 4(c) does not give each customer a right to opt out or limit its recommendation to high-impact decisions.

Tools for explanation, Section 2(f). Section 2(f) names a tool that LFIs "may wish to consider": the Shapley Additive Explanations (SHAP), "to explain the output of models". Section 2(f) names an alternative: "any other well-documented approaches to enhance algorithmic transparency". SHAP is an example. The Guidance Note does not make it mandatory.

Human oversight

Section 7 of the CBUAE Guidance Note gives LFIs recommendations on human oversight and consumer protection.

Meaningful oversight, Section 7(a). Section 7(a) recommends that LFIs make sure that AI and ML systems "operate under meaningful human oversight and judgement". Section 7(a) puts emphasis on these two areas:

• "decisions that have significant implications for consumers"
• "the ongoing selection of, determination as to third party providers of, deployment of and ongoing monitoring and general use of AI".

Oversight models, Section 7(a)(i)–(iii). Section 7(a) tells LFIs that "Human oversight may be exercised through different models". It names three models:

Model	Section	How the AI and the human work
Human-in-the-loop	7(a)(i)	"AI provides recommendations but a human decision maker retains full authority to approve or reject the outcome"
Human-on-the-loop	7(a)(ii)	"the AI works autonomously for routine tasks, while a human monitors outcomes and can intervene where necessary"
Human-out-of-the-loop	7(a)(iii)	"the AI operates without direct human involvement, which should only be utilised for low-risk, non-material processes with appropriate controls in place"

Level of human involvement, Section 7(b). Section 7(b) gives this recommendation: "The level of human involvement should be commensurate with the identified and potential risks posed to a consumer by any AI." Section 7(b) applies to the level of human involvement. It does not change the test events in Section 3(c).

Stop the use of AI, Section 6(f). Section 6(f) gives this recommendation: "LFIs should at all times retain the clear and immediate ability, with human intervention, to cease use of an AI model system, technology or application deployed or utilized." For more about this recommendation, see Incident response and remediation.

Human review and alternatives, Section 7(c). Section 7(c) recommends that consumers can "request human review or explanation" of AI decisions and that customers who do not want AI decisions have alternatives, without a high-impact threshold.

Related information

Section 7(d) of the CBUAE Guidance Note gives recommendations on fair and equitable treatment and on marketing. Section 7(d) also gives recommendations on the disclosures in promotional materials and chatbots: see Consumer treatment and marketing.

In Modulos, UAE Consumer AI uses OFF-21 at the Organization level and MFF-21 at the AI System level. For the Requirements that Modulos maps to Sections 3, 4, and 7, see Requirement mapping. For other Frameworks, see Comparison.

Scope and governance

Scope, definitions, accountability, the AI inventory, and risk rating

Data, models, monitoring, and remediation

Data, model tests, automatic updates, monitoring, and incident response

Consumer redress and third-party AI

Human review, complaints, Article 8, third-party AI, and marketing

Operationalizing in Modulos

Set up OFF-21 and MFF-21, select the scope, and find the Requirements

Sources

• CBUAE Guidance Note — Guidance Note on the Consumer Protection and Responsible Adoption and Use of Artificial Intelligence and Machine Learning by Licensed Financial Institutions in the U.A.E. The Central Bank of the UAE issued it on 11 February 2026. Provisions: Section 1 (High-impact decision), Sections 2(f), 3(a)–(c), 4(a)–(c), 6(c), 6(f), and 7(a)–(d). CBUAE Rulebook.

Disclaimer

This page gives general information. It is not legal advice.