Principles of Generative Artificial Intelligence

Artificial intelligence constitutes the axis of digital transformation around the world, as it drives the wheel of innovation and redefines the standards of efficiency and services across all sectors. Its applications demonstrate a superior ability to analyze big data, automate processes, and make informed decisions, which contributes to raising productivity, reducing costs, and improving quality of life, both at the governmental and private sector levels. Artificial intelligence also facilitates the provision of customized and rapid services and supports evidence-based policymaking, making it a fundamental pillar for building smart cities and sustainable societies.

However, these tremendous capabilities are coupled with major ethical and legal responsibilities. Handling sensitive personal data, the possibility of hidden bias in generative artificial intelligence algorithms, and the legal implications of automated decisions all require the establishment of robust regulatory frameworks. In this context, the Saudi Data and Artificial Intelligence Authority (SDAIA) issued the document of Principles for the Use of Generative Artificial Intelligence for the public and for government entities, in a manner consistent with the regulations and policies in the Kingdom of Saudi Arabia, including regulations and policies related to data governance, privacy, security, intellectual property, and human rights.

Definition of generative artificial intelligence

Generative artificial intelligence, according to the aforementioned principles document, is a machine-learning model capable of generating new examples similar to the training dataset. It operates by interpreting commands provided by users, as it performs tasks that require human cognitive abilities, including responding to verbal or written prompts, learning, and problem-solving. Thus, generative artificial intelligence is considered a subfield of artificial intelligence that can generate new content (including texts, images, sounds, code, video clips, and others). (1)

Accordingly, generative artificial intelligence has become an integral part of several sectors, foremost among them the education and scientific research sector, and regulating its use has become necessary to keep pace with this development. (2)

Principles for the use of generative artificial intelligence for the public and for government entities

The document set out several principles that developers, users, and government entities must adhere to when dealing with generative artificial intelligence at all stages of the life cycle of these tools, in order to maximize their benefits and reduce their risks. These principles are as follows:

1- Integrity and fairness

This principle consists of ensuring that the content generated by artificial intelligence is not biased in a manner that results in negative discrimination against individuals, groups, or specific categories. The document even stipulates the need to ensure that the content does not lack representation of segments of society.

This is achieved by testing artificial intelligence models to ensure that the code and algorithms of artificial intelligence do not contain any bias or discrimination, understanding the data used to train artificial intelligence, and raising awareness of bias and combatting racism and discrimination. (3)

As for employees in government entities, they must learn to write artificial intelligence prompts in a manner that ensures the creation of bias-free content, and government entities must train them for this purpose. (4)

2- Reliability and safety

Reliability means the confidence in the robustness of artificial intelligence systems and how they operate.

Safety refers to the degree of security of artificial intelligence systems and the extent to which they do not represent a danger to society and individuals. This is achieved by designing and developing systems to withstand uncertainty and instability, ensuring that generative artificial intelligence systems have the option to activate human oversight, notifying users of potential reliability challenges with the possibility of verifying the content from other sources, and finally reviewing the content to ensure its accuracy and overall context to prevent the dissemination of misinformation. (5)

Government entities must audit the content generated by artificial intelligence to ensure its accuracy and suitability for context to prevent the dissemination of misinformation. (6)

3- Transparency and explainability

Generative artificial intelligence systems must be built with a high degree of clarity and explainability, with the ability to trace the stages of decision-making and the logic on which those systems base their content.

This is achieved by disclosing to the public when generative artificial intelligence is used when interacting with the public, providing alternative communication channels for users who prefer human interactions, and using distinctive labels to identify content generated by generative artificial intelligence. (7)

4- Accountability and responsibility

Developers and designers of generative artificial intelligence systems bear “ethical responsibility” for the decisions of these systems that may lead to potential risks and negative effects on society, as well as the legal consequences of the content generated by those systems, especially in areas relating to intellectual property violations, personal data, and human rights.

This is achieved by facilitating human intervention in artificial intelligence systems, increasing the quality of the data used in content generation, and complying with relevant laws such as the Personal Data Protection Law issued by Royal Decree No. (M/19) dated 9/2/1443 AH, and laws relating to intellectual and industrial property such as the Law of Patents, Layout-Designs of Integrated Circuits, Plant Varieties, and Industrial Designs issued by Royal Decree No. (M/27) dated 29/5/1425 AH, and the Copyright Protection Law issued by Royal Decree No. (M/41) dated 2/7/1424 AH. (8)

5- Privacy and security

Generative artificial intelligence systems must be developed and used in a manner that ensures the protection of the data they collect to produce their content, and in a manner compatible with cybersecurity requirements, by applying adequate safeguards and security measures when using the systems, strictly applying the provisions of the Personal Data Protection Law, assessing risks resulting from the use of artificial intelligence in accordance with the Principles of Artificial Intelligence Ethics issued by the Saudi Data and Artificial Intelligence Authority, and designing systems in a manner that takes privacy and security mechanisms into account. (9)

Government employees must not input any data classified at restricted level or higher into generative artificial intelligence systems (Restricted, Confidential, and Top Secret) and must verify the privacy policies of these tools, including disclosure and data-sharing clauses. (10)
This principle has great significance in the field of government work due to the damage caused by the use of generative artificial intelligence in making high-risk or highly significant decisions. (11)

6- Humanity and social and environmental benefits

This principle includes the positive role that artificial intelligence must play in developing the skills of individuals and government entities alike, based on a methodology rooted in human rights and enhancing human capabilities. (12)

How does the SDAIA define generative artificial intelligence, and what distinguishes it from other machine-learning models?

According to the source, the Saudi Data and Artificial Intelligence Authority (SDAIA) defines generative artificial intelligence as a machine-learning model that possesses the capability to generate new examples similar to the training dataset it was originally provided.

It is distinguished from other machine-learning models and broader artificial intelligence applications by several key characteristics:

• Content Creation: Unlike models that may focus primarily on analyzing data or automating existing processes, generative AI is a subfield specifically capable of generating entirely new content, including texts, images, sounds, code, and video clips.
• Human Cognitive Abilities: It performs tasks that traditionally require human cognitive abilities, such as learning, problem-solving, and interpreting specific commands or prompts provided by users.
• Interaction Method: It operates by responding directly to verbal or written prompts, translating user input into original outputs that mimic the patterns of its training data.

While broader AI applications are noted for their ability to analyze big data and support evidence-based policymaking, generative AI’s unique value lies in this ability to produce new, synthetic examples rather than just classifying or predicting based on existing data

Which specific Saudi Arabian laws (such as Personal Data Protection and Copyright) must AI developers comply with?

Developers of generative artificial intelligence systems are required to comply with specific Saudi Arabian laws to address the legal and ethical consequences of the content their systems generate. These include:

• Personal Data Protection Law:   Issued by Royal Decree No. (M/19) dated 9/2/1443 AH.
• Copyright Protection Law:   Issued by Royal Decree No. (M/41) dated 2/7/1424 AH.
• Law of Patents, Layout-Designs of Integrated Circuits, Plant Varieties, and Industrial Designs: Issued by Royal Decree No. (M/27) dated 29/5/1425 AH.

Beyond these specific statutes, the SDAIA framework requires that generative AI use remain consistent with broader national regulations and policies concerning data governance, privacy, security, intellectual property, and human rights. Additionally, systems must be developed in accordance with the Principles of Artificial Intelligence Ethics issued by the Saudi Data and Artificial Intelligence Authority.

Overview of Principles of Generative Artificial Intelligence

The text outlines the regulatory framework and ethical guidelines established by the Saudi Data and Artificial Intelligence Authority (SDAIA) for the responsible use of generative technology. It emphasizes that while artificial intelligence acts as a catalyst for global innovation and efficiency, it requires strict adherence to principles like fairness, reliability, and transparency to mitigate risks. The sources detail specific mandates for government entities and the public, focusing on the prevention of algorithmic bias and the protection of sensitive personal data. Furthermore, the document highlights the legal obligations regarding intellectual property and human rights, ensuring that automated systems remain accountable and safe. Ultimately, these principles aim to harmonize technological advancement with societal well-being and national security standards.

---

 

Sources:

1. Paragraph (1) of the Principles of Generative Artificial Intelligence for the Public for the year 2025, issued by the Saudi Data and Artificial Intelligence Authority <SDAIA | Access to SDAIA Publications>.
2. Brittain Shah, Artificial Intelligence and the Future of Education, publisher “Jossey-Bass,” first edition, 2023.
3. Paragraph 1 of item (4) of the Principles of Generative Artificial Intelligence for the Public.
4. Paragraph 1 of item (3) of the Principles of Generative Artificial Intelligence for Government Entities for the year 20205, issued by the Saudi Data and Artificial Intelligence Authority <SDAIA | Access to SDAIA Publications>.
5. Paragraph 2 of item (4) of the Principles of Generative Artificial Intelligence for the Public.
6. Paragraph 2 of item (3) of the Principles of Generative Artificial Intelligence for Government Entities.
7. Paragraph 3 of item (4) of the Principles of Generative Artificial Intelligence for the Public.
8. Paragraph 4 of item (4) of the Principles of Generative Artificial Intelligence for the Public.
9. Paragraph 5 of item (4) of the Principles of Generative Artificial Intelligence for the Public.
10. Paragraph 5 of item (3) of the Principles of Generative Artificial Intelligence for Government Entities.
11. Vincent Straub, Deborah Morgan, Jonathan Bright, and Helen Margetts, The Use of Artificial Intelligence in Government, published in Government Information Quarterly, Vol. (40), Issue (4), 2023.
12. Paragraphs 6 and 7 of item (4) of the Principles of Generative Artificial Intelligence for the Public.