Arabic
info@ifai-uo.org

AI Ethics Guidelines

Responsible Institutional Framework

Artificial Intelligence Ethics Guidelines

A practical framework that helps institutions, developers, and decision-makers design artificial intelligence systems that are safe, fair, transparent, and accountable while protecting human rights, strengthening trust, and translating ethical principles into measurable and actionable practices.

Designed for Institutions and Developers

Explainability
Providing clear and understandable explanations that are appropriate to the nature of the decision and the affected audience.
Fairness
Reducing bias and ensuring fair treatment across different groups and users.
Accountability
Clearly defining human and legal responsibilities throughout the system’s lifecycle.
Transparency
Disclosing the use of intelligent systems and enabling the traceability of important decisions.

Governance Assessment Dashboard

Continuous Assessment

From Principles to
Practice

Trust is established when compliance can be measured, documented, and reviewed throughout the system lifecycle.

Responsibility Clarity
88
Data Quality
78
Explainability
71
Human Oversight
83
Why Do We Need an Ethical Framework?

Trust Is Not an Added Feature, but a Fundamental Requirement for Responsible Use

Intelligent systems may affect employment opportunities, financial services, healthcare, education, security, and access to information. Their evaluation should therefore not be limited to accuracy and speed, but should also consider their impact on individuals, society, and human rights.

Protecting Rights and Interests
Enhancing Institutional Trust
Improving Decision Quality
Governing Principles

Eight Principles for Building Trustworthy AI Systems

The following principles work as an interconnected whole: transparency cannot be achieved without accountability, and fairness cannot be sustained without high-quality data and effective human oversight.

Accountability and Responsibility
A clear owner should be assigned to each system, with responsibilities distributed across management, developers, operators, and technology providers, while ensuring that responsibility is never attributed to the system itself.
Fairness and Non-Discrimination
Outcomes should be tested across different groups, and sources of bias in data, design, or use should be identified and addressed both before and after deployment.
Transparency and Disclosure
Users should be informed when they are interacting with an automated system, with clear information provided about the system’s purpose, limitations, primary data sources, and the entity responsible for it.
Explainability
Explanations should be appropriate to the sensitivity of the decision, clear to non-specialists, and accessible without unnecessary complexity or unjustified cost.
Privacy and Data Governance
Only the minimum necessary data should be collected, with clear rules governing its use, retention, and sharing, alongside appropriate access controls and protection measures.
Safety, Security, and Reliability
The system should be tested against errors, attacks, and misuse, with mechanisms in place for safe shutdown, incident response, and monitoring for performance degradation.
Human Oversight
Humans should retain an active role in reviewing sensitive decisions and be able to stop the system or override its decisions when risks or circumstances require it.
Social Impact and Sustainability
The system’s impact on society, employment, and the environment should be assessed, while innovation should be directed toward delivering meaningful, sustainable, and inclusive benefits across different groups.

How Can Guidelines Be Turned into Institutional Practice?

Commitment begins before the system is developed and continues after deployment through ongoing monitoring, documentation, review, and response to feedback and incidents.

01

Define the Purpose

Define the problem, the expected benefit, and possible non-automated alternatives.

02

Assess the Risks

Analyze potential impacts on rights, safety, and affected groups.

03

Design Safeguards

Establish safeguards for data, security, fairness, and human oversight.

04

Testing and Documentation

Test performance, bias, and robustness, and document decisions and limitations.

05

Responsible Deployment

Inform users and establish clear channels for support, appeals, and escalation.

06

Monitoring and Improvement

Monitor real-world performance, review incidents, and periodically update safeguards.

Individual Rights

The Human Must Remain at the Center of Decision-Making

When automated systems affect individuals, practical safeguards should be in place to enable them to understand the decision, challenge it, request a review, and correct inaccurate information.

Clearly inform individuals when they are interacting with an artificial intelligence system, and explain the general purpose of its use, the entity responsible for it, and the nature of its potential impact.

Provide a context-appropriate explanation that clarifies the key factors influencing the decision, without relying solely on vague or incomprehensible technical responses.

Provide a clear process for challenging an automated decision, supported by meaningful human review with the authority to modify or overturn the decision when necessary.

Enable individuals to request the correction of inaccurate data used in decision-making and, where possible, address the impact of such errors on previous outcomes.

Design preventive measures and rapid incident-response procedures that minimize harm, while clearly defining support channels, accountability, and appropriate remedies or compensation.

Contribute to Building a More Responsible Future for Artificial Intelligence

Ethical guidelines evolve alongside technology and the contexts in which it is used. We welcome feedback from researchers, developers, institutions, and the wider community to help improve this framework and expand its application.
Share Your Feedback