United Arab Emirates - AI Ethics Self-Assessment Tool

Overview

The AI Ethics Self‑Assessment Tool (beta) is a voluntary self‑assessment instrument published by UAE authorities to operationalise the country and city level AI ethics principles for individual AI systems. The tool is available online and described on official portals maintained by Digital Dubai and the UAE Artificial Intelligence Office. It enables organisations to classify an AI system’s decision impact (non‑significant, significant, critical), run structured self‑assessments across fairness, accountability, transparency and explainability domains, and receive a results page showing scorecards and flagged gaps. The tool is explicitly marked as beta to invite feedback and use‑case submissions; it is published under a Creative Commons Attribution 4.0 International Licence to encourage reuse. The tool is intended for AI developers, operators and public sector procurement teams and is non‑audited during the beta period; it is therefore guidance‑driven rather than a compliance regime enforced by regulatory sanctions. Official presentations and descriptions appear on the Digital Dubai platform and are mirrored on the UAE AI Office site. See the online tool at Digital Dubai — AI System Ethics Self‑Assessment Tool.

Definitions

Key terms used by the tool align with common governance vocabulary: “AI system” refers to a technical system that automates or augments decision‑making; “developer” denotes the organisation or team creating the model or pipeline; “operator” denotes the organisation operating or deploying the system; “decision significance” classifies expected impact on individuals or groups as non‑significant, significant or critical; “mitigation measures” are controls (technical, process, governance) applied to reduce identified ethical or legal risks; and “ethics performance score” is the derived output indicating how the AI system measures against recommended guidelines. The tool uses recommendation strengths (‘should’ = high, ‘should consider’ = moderate) to weight scoring outcomes and requires explanations when a guideline is marked Not Applicable.

Governance and Institutional Framework

The tool is situated within Dubai’s and the UAE’s broader ethics and AI governance ecosystem. Digital Dubai (formerly Smart Dubai) created city‑level AI Ethics Principles & Guidelines and the self‑assessment toolkit as a practical instrument for public and private actors; the Executive Council of Dubai has advised public entities to use the toolkit when implementing AI. Separately, the UAE Artificial Intelligence Office and federal portals publish complementary resources and position papers (including an AI Ethics Guide and the UAE Charter for the Development & Use of AI). Institutional custodianship is shared: in Dubai, Digital Dubai manages the toolkit and consults its AI Ethics Advisory Board; at the federal level, the AI Office curates national publications and resources. The tool’s governance model emphasises voluntary adoption, iterative improvement through stakeholder feedback, and reuse via an open licence; it is therefore primarily an advisory governance instrument, designed to be integrated into existing risk and procurement frameworks rather than to operate as a standalone regulatory instrument. See the Digital Dubai description at Digital Dubai — Self‑Assessment and federal materials at UAE AI Office — Publications.

Key Focus Areas

The tool concentrates on operationalising principles across four core ethical dimensions: fairness (data representativeness, bias detection and mitigation), accountability (roles, recordkeeping, audit trails, vendor governance), transparency (user notices, pre‑deployment disclosure, documentation of intended use and limitations) and explainability (methods for interpreting model outputs appropriate to decision significance). It also addresses lifecycle topics: classification of decision significance, data‑use risks, third‑party model and methodology risks, historic bias, technical bias, and ongoing monitoring. The self‑assessment comprises modular question sets and recommended mitigations; stronger recommendations have greater influence on the final ethics performance score. The tool encourages documentation (e.g., datasheets, model cards), design of human‑in‑the‑loop processes for high‑impact decisions, testing protocols and periodic re‑assessment to capture model drift or environmental change. The guidance is interoperable with sectoral obligations (for example, healthcare or financial compliance requirements) and complementary to data protection obligations under UAE Personal Data Protection Law. The document and tool provide example mitigation strategies and emphasise the need for justification when classifying a guideline as Not Applicable.

Implementation Framework

Organisations are instructed to begin at the General Info & Classification stage, documenting system purpose, stakeholders and decision significance. The self‑assessment then proceeds through the four ethical modules, requiring self‑judged performance levels and selection/description of mitigation measures. Results generate a scorecard and highlight priority gaps. Recommended implementation practices include: embedding the tool into procurement and development lifecycles, maintaining audit logs and documentation for each assessment, instituting internal review committees for significant systems, conducting technical bias and robustness testing prior to deployment, and scheduling regular reassessments. Because the tool is permissive, organisations are expected to adopt the elements appropriate to their risk profile; public entities in Dubai were specifically encouraged to use the toolkit per Smart Dubai communications. The tool’s open licence allows organisations to adapt the questionnaires into internal workflows and to share anonymised use cases to improve benchmarking.

Monitoring and Evaluation

Monitoring under the beta tool is organisation‑led: the tool provides a structure for internal monitoring through periodic self‑assessments, documentation of mitigation actions and the results page which identifies remaining gaps. Users are invited to submit feedback and case studies through the Digital Dubai feedback form to enable iterative improvement and potential benchmarking across organisations. There is no central audit function in the beta phase; however, the documentation generated by the tool can be used in future for procurement checks, internal or external audits, or to evidence due diligence in the event of regulatory scrutiny under other UAE laws (e.g., data protection, sector safety standards). Metrics for evaluation recommended by the tool include reduction in identified gap counts, improvements in weighted ethics score, frequency of reassessment, and number of implemented mitigation measures.

Penalties, Liability, and Appeals

The AI Ethics Self‑Assessment Tool itself imposes no direct penalties because it is voluntary and non‑binding in its beta form. Liability and penalties for AI‑related harms remain governed by UAE substantive laws and sectoral regulations (for example, personal data protection statutes, healthcare safety regulations, or criminal provisions for misuse). The tool advises organisations to maintain documentation that can support legal defences and appeals in the event of disputes. In practice, documented internal assessments may mitigate enforcement outcomes if they demonstrate good‑faith risk management, but they do not substitute for legal compliance where statutory obligations exist. Entities should therefore use the tool alongside formal legal counsel and sectoral regulatory compliance processes.

Relationship to Other Instruments

The tool complements Dubai’s AI Ethics Principles & Guidelines (and earlier Smart Dubai materials) and the UAE AI Office’s national ethics publications and charter. It is intended to translate high‑level principles into system‑level checks and to interoperate with data protection impact assessments and sectoral risk assessments. The tool’s open licence and the Digital Dubai feedback mechanism are meant to produce convergence between city and federal practices and to provide practical guidance that can be embedded into procurement and internal governance. See the Digital Dubai principles at Digital Dubai — AI Ethics Principles & Guidelines.

International Alignment

The tool draws on international best practice and aligns conceptually with OECD, ISO and other international AI ethics recommendations by foregrounding fairness, accountability, transparency and safety. Its modular approach (classification by impact, weighted recommendations and lifecycle assessments) resembles similar governance tools published by international actors (e.g., ICO’s AI and Data Protection Risk Toolkit, NIST risk management resources). The open licence supports international reuse and cross‑jurisdictional benchmarking. The tool is positioned to be interoperable with standards such as ISO/IEC AI trustworthiness frameworks and EU/UK instruments, facilitating multinational organisations’ alignment across jurisdictions.

Implementation Timeline

Sources and References