Riyadh hosts the Global Forum on the Ethics of Artificial Intelligence at a time when the world is moving from developing the machine to regulating its power and protecting people from its decisions

Prepared and Analyzed by | Strategic Media Department – BETH Agency
Supervised by: Abdullah Al-Omirah

Riyadh | BETH

The global question is no longer: What can artificial intelligence do?

AI has moved beyond writing, translation, and image analysis. It has entered medicine, education, media, justice, markets, and government services, and now participates in decisions that may determine who receives a job, a loan, medical treatment, an educational opportunity, or even who is subjected to surveillance and investigation.

The more urgent question has therefore become: What should artificial intelligence be allowed to do, and who bears responsibility when it makes a mistake?

This is where the ethics of artificial intelligence begins. It is not an attempt to give the machine a conscience, because a machine possesses neither conscience nor intent nor a sense of guilt. Rather, it is a system of rules and standards that governs the decisions of the people and institutions that design, train, and use it.

The algorithm does not bear moral responsibility. Responsibility lies with those who gave it the data and authority, and then adopted its decision.

What Does Ethics Mean Here?

The ethics of artificial intelligence means ensuring that systems are fair, safe, and transparent; that they respect privacy and human dignity; that they do not discriminate among people; and that they do not replace human beings in decisions affecting their rights and lives.

Technical accuracy alone is not enough. A system may provide a statistically correct answer that is nevertheless unjust to a person or contrary to the law.

An automated system may reject a job applicant because it learned from past records that favored a particular group. It may give a patient lower treatment priority because its training data did not adequately represent that patient’s condition. It may produce convincing but fabricated media content or use people’s faces and voices without their consent.

Ethics therefore does not examine the system’s capabilities alone. It also examines the purpose for which it is used, the quality of its data, who is affected by its decisions, and whether those decisions can be challenged and corrected.

People First

UNESCO’s Recommendation on the Ethics of Artificial Intelligence, adopted by Member States in 2021, places human dignity and human rights at the center of governance.

The Recommendation is based on key values that include fairness, diversity and inclusion, transparency, environmental sustainability, and human oversight. It also affirms that ultimate responsibility cannot be transferred from a human being to an automated system.

This means that the common phrase “the system made the decision” cannot serve as a justification when harm occurs.

The system did not choose its objective, collect its own data, or define the limits of its authority. At every stage, an entity or individual made a decision, and that party must be identifiable and accountable.

Bias That Learns

Artificial intelligence bias does not necessarily begin with the machine. It may already exist within society and the data.

If a recruitment system is trained on previous human decisions that included discrimination, it may learn and reproduce that discrimination more quickly and on a wider scale, even without displaying explicitly racist or biased language.

Bias may also result from underrepresentation. When systems are trained on data that does not adequately represent women, older people, people with disabilities, or certain dialects and languages, they become less accurate in dealing with them.

The danger is that a biased human decision may harm one person, while a biased system can repeat the same mistake thousands of times and turn it into an institutional procedure that appears neutral merely because it was issued by a machine.

Fairness therefore requires testing outcomes across different groups, rather than relying solely on the system’s overall average accuracy.

Where Did the Answer Come From?

Trust in artificial intelligence depends on the ability to understand how it reached a decision.

Transparency does not mean exposing commercial secrets or publishing millions of lines of code. It means informing people that they are dealing with an automated system, explaining the main factors influencing the decision, identifying the source of the data and the limits of accuracy, and providing a path for appeal and human review.

UNESCO emphasizes that people whose rights are affected by an automated decision must be able to understand the reasons behind it and request an explanation and correction.

This becomes particularly important when the decision concerns employment, credit, insurance, education, government services, or criminal justice.

The greater a decision’s impact on a person’s life, the stronger that person’s right to understand and challenge it.

Privacy Is Not Free Fuel

Artificial intelligence requires vast quantities of data, but technical need does not grant an unlimited right to collect people’s lives.

Images, voices, locations, health records, messages, and purchasing habits are not raw materials without an owner. They are extensions of a person’s identity and privacy.

Protection begins with the question of necessity: Does the service genuinely need this data? It then extends to consent, purpose, retention period, who may access the information, and how it is protected from leaks or reuse.

An individual may consent to the use of data for a specific service, but that does not mean consenting to its use in training a commercial model or sharing it with third parties.

The ethical rule here is clear: Do not collect everything you can. Collect only what you need, for a purpose known to its owner.

Human Oversight

Keeping a human being within the decision-making process may appear reassuring, but it can become a merely symbolic procedure if the employee automatically approves the system’s recommendations.

Genuine oversight requires reviewers to have the time, knowledge, and authority to reject the machine’s decision, as well as an understanding of its limitations and the reasons it may be wrong.

It is also necessary to identify the decisions that must never be fully delegated to artificial intelligence, particularly those involving life and death, punishment, fundamental rights, and critical diagnoses.

UNESCO’s Recommendation states that artificial intelligence cannot replace ultimate human responsibility and that, as a general rule, life-and-death decisions should not be delegated to automated systems.

Who Pays for the Mistake?

When an artificial intelligence system makes a mistake, responsibility may be distributed among the model developer, the data provider, the entity that purchased it, the institution that used it, and the employee who adopted its outcome.

This distribution can cause responsibility to disappear among all parties.

Governance therefore requires clear answers from the outset:

• Who authorized the system’s use?
• Who tested it before deployment?
• Who monitors its performance after launch?
• Who receives complaints?
• Who stops it when harm appears?
• Who compensates the affected person?

Ethics that fails to identify the responsible party does not prevent error. It merely describes it after it occurs.

Risks Are Not Equal

A system that recommends a music playlist does not require the same level of oversight as one that selects job applicants or assists in diagnosing diseases.

Global governance is therefore moving toward classifying systems according to their level of risk.

The European Union’s AI Act divides uses into levels beginning with low-risk applications, extending to high-risk systems, and ending with unacceptable uses that are prohibited because they threaten safety or rights.

The idea is not to slow every innovation, but to increase requirements as the scale of potential harm rises.

A high-risk system requires reliable data, documentation, independent testing, decision logs, human oversight, and mechanisms for reporting incidents and submitting appeals.

Media and Truth

Media is one of the fields most exposed to the ethical challenges of artificial intelligence.

AI systems can produce news stories, images, audio, and video within seconds, imitate real people, and flood platforms with content that audiences struggle to distinguish from authentic material.

The danger is not limited to false news. It also extends to weakening trust in everything. When images and voices can be easily fabricated, a guilty person can claim that authentic evidence is fake, while fabricated material can appear to be genuine evidence.

Media organizations need clear rules that include verifying content, disclosing generated material, refraining from inventing quotations, protecting images and voices, and ensuring that editorial responsibility remains with an identifiable journalist.

Artificial intelligence can assist journalists with research, analysis, and translation, but it does not bear responsibility for publication, face the person harmed, or consistently distinguish between accurate information and a sentence that merely sounds convincing.

Education and the Mind

Artificial intelligence can personalize learning, explain material in different ways, and help teachers analyze students’ needs.

But it may also transform students from seekers of knowledge into consumers of ready-made answers and weaken their ability to read, write, and analyze when it is used as a substitute for thinking.

Monitoring students and analyzing their behavior, faces, and emotions also raise risks related to privacy and misclassification.

The question in education should therefore not be: Did the student use artificial intelligence?

It should be: How was it used, what part did the student complete through independent thought, and did the student become more capable of understanding or merely more capable of concealing a lack of understanding?

Healthcare: A Sensitive Decision

Artificial intelligence can identify patterns that a physician may not notice, rapidly analyze images and records, and help predict risks.

But an error in healthcare does not merely produce a poor piece of text. It may delay treatment or recommend an inappropriate procedure.

Medical systems therefore require data that represents different groups, clinical testing, continuous monitoring, and strict protection of health information, while the decision and responsibility remain with the specialist.

Patients must know when artificial intelligence has been used to assess their condition, the limits of its role, and who can review the result.

Working Under the Algorithm

Artificial intelligence may help companies improve productivity and identify skills, but it can also monitor, assess, and exclude employees according to criteria they do not understand.

The problem emerges when efficiency becomes permanent surveillance, when workers are required to prove their superiority over a machine that never sleeps, or when job applicants are rejected without knowing why.

The workplace must guarantee employees the right to know when they are being assessed automatically, what data is being used, whether they can appeal, and how digital indicators are prevented from becoming a final judgment on a human being.

Arabic Is Not Peripheral

Ethics is also connected to language and culture.

Systems trained primarily on content from particular languages and cultures may understand Arabic, its dialects, and its contexts less accurately, or transfer cultural assumptions that do not reflect the society in which they are used.

A shortage of reliable Arabic content may lead to lower-quality answers or bias against local names, dialects, and concepts.

It is therefore not enough to import and translate models. High-quality Arabic data must be developed, local diversity represented, and models tested within the context in which they will operate.

Linguistic justice is part of digital justice. Those absent from the data of the future may not be properly understood by its decision-making systems.

A Cost the User Does Not See

An artificial intelligence response may appear immaterial, but it is produced inside data centers that consume energy, water, and chips.

As models expand, sustainability becomes part of their ethics. Is the benefit proportional to the resources consumed? Is clean energy being used? Is the system’s environmental footprint measured during training and operation?

Ethics here does not call for halting development. It calls for developing more efficient models, selecting the appropriate model size for the task, and avoiding the consumption of vast resources to perform work that simpler tools could accomplish.

From Principles to Operation

The global problem is not a shortage of ethical documents. UNESCO, the United Nations, governments, and many technology institutions have issued numerous principles.

The problem is turning words into procedures within institutions.

Ethical operation requires a continuous cycle that begins before a system is purchased or developed and includes:

• Defining the purpose and necessity.
• Assessing the impact on rights and privacy.
• Examining data sources and quality.
• Testing bias, accuracy, and safety.
• Defining responsibilities and authorities.
• Providing review and appeal mechanisms.
• Documenting decisions and incidents.
• Monitoring the system after deployment.
• Stopping or modifying it when it exceeds acceptable limits.

The U.S. National Institute of Standards and Technology’s AI Risk Management Framework proposes managing artificial intelligence risks through four interconnected functions: governance, understanding the context, measurement, and risk management. Testing before launch is not enough for a system whose performance may change as its data and uses evolve.

Saudi Arabia Builds the Framework

Saudi Arabia’s interest in the ethics of artificial intelligence does not begin with hosting the Global Forum.

The Saudi Data and Artificial Intelligence Authority has established principles and guidelines for artificial intelligence ethics and introduced a self-assessment tool that allows government, private-sector, and nonprofit entities to measure how closely their models comply with ethical standards.  

The importance of the assessment lies in moving ethics from a general declaration to questions that can be examined: How was the data collected? Was bias tested? Who reviews the results? Can an affected person appeal? What levels of transparency and safety are provided?

The next challenge, however, is to standardize practices across sectors, train specialists in algorithmic auditing, and establish stricter requirements for systems used in healthcare, education, employment, and public services.

Riyadh Hosts the World

Riyadh will host the fourth UNESCO Global Forum on the Ethics of Artificial Intelligence from September 14 to 17, 2026. The event is jointly organized by UNESCO and Saudi Arabia, represented by the Saudi Data and Artificial Intelligence Authority and the International Center for AI Research and Ethics.

The Forum will be held under the theme “Transforming Global Cooperation for Ethical AI Governance” and will bring together governments, the private sector, researchers, civil society, and international organizations.

Its topics include supervision and accountability, institutional readiness, public administration, education and healthcare, equality and inclusion, cultural and linguistic diversity, AI safety, energy and sustainability, and international cooperation.

Riyadh’s role is not limited to hosting the Forum. Its selection reflects the Kingdom’s transition from investing in artificial intelligence technologies to participating in shaping the rules governing their use worldwide.

The Forum is particularly significant because it will be held alongside the Global AI Summit as part of Saudi Arabia’s Global AI Week, bringing together in one city those who develop the technology and those who question its limits and responsibilities.

Cabinet Attention

During the session chaired by Custodian of the Two Holy Mosques King Salman bin Abdulaziz, the Cabinet affirmed that Riyadh’s hosting of the fourth edition of the Forum reflects the Kingdom’s advanced position in supporting multilateral global dialogue on emerging technologies.

It also described the event as an embodiment of Saudi Arabia’s commitment to promoting the responsible and ethical use of artificial intelligence, contributing to a more inclusive and sustainable digital future.

The Cabinet’s reference gives the subject a significance that extends beyond organizing an international conference. It places the ethics of artificial intelligence within the state’s path toward building a technology-driven economy, protecting society, and participating in shaping the global governance of emerging technologies.

BETH Analysis

The next competition will not be between the country that possesses the most powerful model and the country that imposes the greatest number of restrictions.

The real competition will be between those who develop artificial intelligence rapidly without earning public trust and those capable of combining innovation, trust, and responsibility.

Unregulated systems may achieve rapid gains, but they also produce errors, legal disputes, and social fear that slow their adoption. Good governance does not obstruct innovation; it creates an environment in which institutions and citizens can trust it.

Ethics does not begin when a system makes a mistake, but before it is built: when the problem is selected, the data is collected, and it is determined who holds the decision and who may challenge it.

The danger is not merely that the machine may become more intelligent than the human being, but that the human being may surrender decisions to it and then claim no longer to be responsible for them.

Another Perspective

Artificial intelligence does not need a conscience, because it will feel no remorse if it wrongs a human being.

The conscience is needed by the designer who trains it, the institution that uses it, the official who hands it the decision, and the society that allows efficiency to take precedence over dignity.

The closer the algorithm comes to a person’s life, the closer the human being must remain to its decision.

The future will not be measured only by what the machine was able to do, but by the wisdom human beings possessed to prevent it from doing what it should not.