Docs

Appearance

Frequently asked questions about the UAE AI Ethics Principles and Guidelines ​

Use this FAQ to examine the UAE AI Ethics Principles and Guidelines and select a Modulos Framework level.

Disclaimer

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

Are the UAE AI Ethics Principles and Guidelines legally binding? ​

No. The UAE AI Ethics Principles and Guidelines (DEC 2022) are "non-binding" (§ 1.1). Section 1.2 gives "non-mandatory guidelines for achieving the ethical design and deployment of AI systems". The guidelines are "part of a toolkit of self-governing nature" (§ 1.6).

All the guidelines are non-binding:

  • Some parts of the guidelines use "must". These parts are also non-binding.
  • This item is about these guidelines only. It does not tell which other UAE laws apply to an AI system.

See Non-binding guidelines.

Which AI systems do the UAE AI Ethics Guidelines apply to? ​

The UAE AI Ethics Principles and Guidelines apply only to AI systems that "make or inform" significant decisions, in the public and private sectors (§ 1.2).

  • A significant decision is "either individually significant or is part of a set of significant-at-scale decisions" (§ 2.11).
  • An individually significant decision has "the potential for significant impact on at least one individual’s circumstances, behavior or choices", or "legal or similarly significant effects on him or her" (§ 2.8).
  • A set of significant-at-scale decisions comes from "the same system or organization" and, together, has "significant impact on society as a whole or groups within it" (§ 2.10). "The decisions need not be individually significant" (§ 2.10).
  • Critical decisions: individually significant decisions, with "a very large impact" or "especially high stakes" (§ 2.6).
  • AI system: a "product, service, process or decision-making methodology whose operation or outcome is materially influenced by artificially intelligent functional units" (§ 2.4). The result can come from AI and other parts of the system (§ 2.4, Note 1).

See decision tests and AI systems.

Can an organization be both an AI developer and an AI operator? ​

Yes. In the UAE AI Ethics Principles and Guidelines, one organization can be an AI developer organization and an AI operator organization (§ 2.2, Note 2).

One test is sufficient for each type:

  • AI developer: "determines the purpose", "designs", "builds", or "technical maintenance or tuning" (§ 2.1), for use or to "sell it on or give it away".
  • AI operator: use in "operations, backroom processes or decision-making", a "service to an AI subject", "business owner of an AI system", "procures and treats data for use in an AI system", or "evaluates the use case for an AI system and decides whether to proceed" (§ 2.2). This includes systems "developed in-house or procured".

These types are different from the Modulos Organization level. See the organization tests.

What are the eight UAE AI Ethics principles? ​

The UAE AI Ethics Principles and Guidelines contain "eight principles of AI ethics" (§ 1.1). Section 3 gives them in this sequence:

  1. "Fairness" (§ 3.1)
  2. "Accountable AI" (§ 3.2)
  3. "Transparent AI" (§ 3.3)
  4. "Explainable AI" (§ 3.4)
  5. "Robust, Safe and Secure AI" (§ 3.5)
  6. "Human Centered AI" (§ 3.6)
  7. "Sustainable and Environmentally Friendly AI" (§ 3.7)
  8. "Privacy Preserving AI" (§ 3.8)

These principles come from the guidelines of DEC 2022. See The eight principles.

Are the federal UAE AI Ethics Guidelines the same as the Dubai guidelines or the UAE AI Charter? ​

No. The UAE AI Ethics Principles and Guidelines, Dubai guidelines and UAE Charter are different:

  • The UAE AI Ethics Principles and Guidelines (DEC 2022), from the UAE AI Office. MFF-4 and OFF-4 in Modulos come from these guidelines.
  • The AI Ethics Principles & Guidelines of Digital Dubai, with the date 01 February 2019. These guidelines are different from the UAE AI Office guidelines. MFF-4 and OFF-4 do not come from these guidelines.
  • The UAE Charter for the Development and Use of Artificial Intelligence. This Charter is different from the UAE AI Office guidelines.

See Sources.

Can people challenge or opt out of AI decisions under the UAE guidelines? ​

The non-binding UAE AI Ethics Principles and Guidelines recommend procedures to challenge significant decisions. The guidelines also recommend that AI subjects can opt out of significant automated decisions "concerning them" "where appropriate" (§ 3.2.2.4).

Section 3.2.2.4 gives AI operators these guidelines:

  • AI operators who use AI systems to "inform significant decisions" "should" give "affected AI subjects" procedures to challenge "a specific decision concerning them" (a).
  • AI operators "should" tell "affected AI subjects" about these procedures and make the procedures "convenient and user-friendly" (c).
  • AI operators "should consider" these procedures for "non-significant decisions" (b).
  • AI operators "should consider" human case evaluators to examine challenges and, "when appropriate, overturn the challenged decision" (d).
  • AI operators "should consider" an opt-out mechanism for significant automated decisions (e).

See Principle 2.

When do the UAE AI Ethics Guidelines recommend an external audit or human review? ​

The UAE AI Ethics Principles and Guidelines recommend external audits for critical decisions and "validation and verification via a human operator" for critical decisions that are life and death decisions (§ 3.2.2.3).

  • If critical decisions are "of civic interest", "public release of the results of the audit should be considered" (§ 3.2.2.3(b)).
  • AI operators "should consider" human case evaluators for challenges (§ 3.2.2.4(d)).
  • Systems that "may directly impact people’s lives in a significant way": the systems "should be able to be overridden" or have "their decisions reversed by designated individuals" (§ 3.5.2.4(d)–(e)).
  • Decisions "related to lethal force, life and death" "should not be delegated to AI systems" (§ 3.6.2.3(a)).
  • If "explainability is not possible given available technology", AI operators "should consider" other methods. The guidelines do not recommend checks by a person only because of this condition (§ 3.4.2.2(b)).

Do the UAE guidelines require AI explanations or public source code? ​

No. The non-binding UAE AI Ethics Principles and Guidelines recommend explanations for decisions and methods with "a significant effect on individuals", as "permitted by available technology" (§ 3.4.1).

Section 3.4 gives AI operators these guidelines:

  • AI operators "should consider" a "means to request explanations" for "affected AI subjects" about "specific significant decisions" (§ 3.4.2.2). This applies "to the extent possible given the state of present research and the choice of model".
  • If explanations are available, they "should be easily and quickly accessible, free of charge and user-friendly" (§ 3.4.2.2(c)).
  • AI operators "should consider" other methods if "explainability is not possible given available technology" (§ 3.4.2.2(b)).
  • AI operators "could consider" an explanation of how the AI system operates for "affected AI subjects" (§ 3.4.2.1).
  • AI operators "should" keep necessary documentation about how the algorithm operates (§ 3.4.2.1(c)).
  • For "non-sensitive public sector use cases designed for the common good", AI operators "could consider" source code with explanations, "either publicly or upon request" (§ 3.4.2.1(b)). This guideline applies only if the risk of "gaming the system" is low.

See Principle 4.

Are privacy impact assessments only for high-risk AI systems? ​

No. The UAE AI Ethics Principles and Guidelines set no high-risk condition for privacy impact assessments. The guidelines apply only to AI systems that "make or inform" significant decisions (§ 1.2).

Section 3.8 gives this information:

  • AI systems "should" use privacy by design (§ 3.8.1).
  • AI algorithms "should" have "adequate privacy impact assessments" where possible (§ 3.8.1).
  • § 3.8.2.2(f): "Algorithmic systems require adequate privacy impact assessments, which also include societal and ethical considerations".
  • AI systems "should respect privacy" with "the minimum intrusion necessary" (§ 3.8.1).
  • The guidelines recommend "specific responsibilities and role for Data Protection Officers" (§ 3.8.2.2(b)).

These are non-binding guidelines, not a data protection law.

See Principle 8.

Does sustainable AI mean reducing cloud costs? ​

No. The sustainability principle of the UAE AI Ethics Principles and Guidelines is about environmental impact, "resource usage and energy consumption" (§ 3.7.1).

Section 3.7 gives this information:

  • Checks of "development, deployment and use" "should" examine "resource usage and energy consumption" (§ 3.7.1).
  • "AI model development" and commissioning "should therefore seek to balance" "technical performance with energy consumption and environmental impact" (§ 3.7.2.1(b)).
  • Checks of environmental impact "should" include the "AI lifecycle", with "energy consumption costs", "CO2 emissions" and "cloud compute costs" (§ 3.7.2.1(c)).
  • Where possible, "model development and commissioning" "should" make "carbon impacts a core consideration alongside functional and business requirements" (§ 3.7.2.1(d)).
  • The guidelines recommend that "Smaller models should be considered", with "fewer compute cycles" to balance "financial and performance costs with the end performance of the model" (§ 3.7.2.1(e)).
  • The guidelines recommend that "Carbon awareness should be considered" to select the "time and location for energy use" (§ 3.7.2.1(f)).

See Principle 7.

Which UAE AI Ethics Framework should I use in Modulos? ​

Use the two UAE AI Ethics Frameworks together: MFF-4 for each AI system in scope and OFF-4 for Organization work.

  • Check scope for significant decisions and the two types of organization first (§§ 1.2, 2.1–2.2). One organization can be of the two types. These types are different from Modulos levels.
  • Add MFF-4 to the AI System record of each AI system in scope.
  • Add OFF-4 to your organization AI System. Contact Modulos if your organization does not have one.
  • Modulos added the "AI Ethics Role" tag for scope selection by type of organization. Each Requirement has the two values. Thus, the tag removes no Requirement for the two types.

See Recommended AI System structure.

Does completing the mapped work prove compliance with the UAE AI Ethics Guidelines? ​

No. For the UAE AI Ethics Principles and Guidelines, complete Controls do not show that a Requirement is Fulfilled. Complete Controls also do not show that your organization follows the guidelines. The guidelines are also non-binding (§ 1.1).

In Modulos:

  • The Requirement Owner examines the completed Controls and can mark the Requirement as Fulfilled.
  • A Fulfilled Requirement does not show that all recommended work is complete. Compare your work with the guidelines.
  • Check the applicable guidelines for external audits (§ 3.2.2.3), information for AI subjects (§ 3.3.2.2), and human control (§ 3.6.2.3).

See Complete the work.