Generative Artificial Intelligence Guidelines for Public
Saudi Arabia · Guideline · Saudi Data and Artificial Intelligence Authority (SDAIA)
Our summary
These guidelines provide non-binding recommendations for developers and users of generative AI systems in Saudi Arabia to promote responsible adoption while managing associated risks. They establish seven core principles - fairness, reliability and safety, transparency, accountability, privacy and security, humanity, and social and environmental benefits - that should guide the lifecycle of GenAI systems. The guidance applies to all stakeholders including public and private entities, researchers, and individuals working with or affected by GenAI technologies.
Written by the Index in our own words. It is not legal advice; the official text is the authority.
Inside this policy
Every section of this policy in order. Our notes summarise each one in plain language and do not quote the official text, which is the authority.
Drafted by the Index An automated summary of the official text, not yet reviewed by a person.Reviewed Checked by a person against the official text.
10 sections
Jump to a chapter
1 Definitions
Defines generative artificial intelligence as machine learning models that create new content similar to training data by interpreting user commands. Defines users as any natural or legal persons consuming services from genAI systems, and developers as those creating genAI systems, including data scientists and researchers.
2 Scope of this Document
This guidance applies to all stakeholders in Saudi Arabia who design, develop, deploy, use, or are affected by genAI systems. The document aims to promote responsible development and use of genAI while building trust and preventing negative outcomes.
3 Benefits of Generative AI
5 sections
This section outlines how genAI is expected to benefit various sectors in Saudi Arabia through five main categories: efficiency and productivity gains, improved communication and collaboration, better decision-making, enhanced accessibility for people with disabilities, and elevated quality of public services.
3.1 Increased Efficiency and Productivity
GenAI can handle routine human tasks such as summarizing documents, transcribing content, or generating images and videos, allowing people to concentrate on more strategic and complex work.
3.2 Improved Communication and Collaboration
GenAI deployment supports content creation, digital avatars, video conferencing optimization, and virtual collaboration environments that strengthen communication and cooperation among users.
3.3 Informed Decision-Making
Generative AI can process and consolidate data from multiple formats, and simulate potential outcomes to support strategic choices. Applied correctly, this technology enables a fact-based decision-making methodology relevant to fields like emergency response, medical treatment, and academic institutions.
3.4 Enhanced Accessibility and Inclusion
GenAI can adapt content for individuals with disabilities to ensure universal accessibility, and generate materials in local dialects and cultural contexts to foster communication across diverse populations.
3.5 Elevated Public Service Quality
Generative AI can improve public service quality and citizen satisfaction by personalizing how services are delivered and increasing their accessibility. It applies across multiple public domains, including medical services, schooling, social programs, and court services.
4 Principles of Generative AI
Developers and users of generative AI tools should follow SDAIA ethics principles across the system's lifecycle, applying to public, private, non-profit entities and individuals: the Fairness principle requires eliminating bias and discrimination through careful model testing, understanding training data origins and contents, and enhancing knowledge of bias and ethics; the Reliability and Safety principle ensures systems work as intended and do not harm society through robust design, assessment standards, human oversight triggers, identification of AI-generated content, and verification of factual accuracy; the Transparency and Explainability principle builds public trust by clearly communicating AI use, notifying users of AI-generated content, offering non-automated channels, and using watermarks to identify AI content; the Accountability and Responsibility principle holds designers, vendors, developers, and others ethically responsible by requiring proper data acquisition and processing, data quality checks, responsible model building, compliance with data protection and intellectual property laws, and legal consultation; the Privacy and Security principle protects data through adequate security measures, data protection principles, risk assessments, privacy by design, and continuous review of impact assessments; the Humanity principle requires building AI systems using ethical methodology based on human rights and cultural values, without designing systems to deceive or manipulate behavior, instead adopting human-centric approaches; and the Social and Environmental Benefits principle focuses on positive impacts that benefit individuals and communities through sustainable goals, empowering progress while preventing harm to human beings and the environment.
5 Generative AI Risks & potential Mitigations
Organizations must train employees that existing legal requirements remain binding when using generative AI, covering data governance, cybersecurity, data classification, personal data protection, and intellectual property rights; mitigate risks from deepfakes and misrepresentation through watermarking, Know Your Customer verification, output verification, and digital literacy training; address safety threats through content moderation, filtering systems, and training dataset filtering; and manage risks from inaccurate or fabricated information by estimating accuracy, citing sources, displaying warning messages for low confidence, and embedding watermarks. Users bear responsibility for verifying and fact-checking generated content and should stay informed of service updates and regularly verify outputs especially in critical applications, while developers should annotate AI-generated code and build expertise in understanding generative AI. Scientists publishing models with potential safety risks should require user verification, service providers should offer consent options for data training use and ability to remove prompt history upon request, and organizations should address risks to certification through assessment redesign, staff and student training on responsible use, collaborative development of clear usage guidelines, and specialized training in effective prompt crafting.
Controls
What this policy requires, forbids, grants or penalises, one control at a time and in our own words. Each control links to the official text, which is the authority. This is not legal advice.
Drafted by the Index Drafted automatically from the official text, not yet reviewed by a person.Reviewed Checked by a person against the official text.
Themes covered 10 of 35
Choose a theme to see only its controls. About the themes
Risk and scope 1 of 4
- Risk classification, not covered
- Prohibited practices, not covered
- Impact assessment, not covered
Data and privacy 1 of 4
- Data subject rights and consent, not covered
- Cross-border data transfer, not covered
- Copyright and training data, not covered
Transparency and records 2 of 4
- Labelling of synthetic content, not covered
- Record keeping and logging, not covered
Technical safeguards 3 of 4
- Human oversight, not covered
Assurance and lifecycle 0 of 4
- Conformity assessment and certification, not covered
- Registration and notification, not covered
- Incident and breach reporting, not covered
- Post-market monitoring and corrective action, not covered
Rights and protected groups 1 of 4
- Automated decision-making rights, not covered
- Biometric identification and surveillance, not covered
- Whistleblowing, complaints and redress, not covered
Roles and organisation 2 of 6
- Supply chain and provider-deployer duties, not covered
- General purpose and foundation models, not covered
- Accountability and governance roles, not covered
- Innovation support and sandboxes, not covered
Enforcement and sectors 0 of 5
- Enforcement, penalties and regulators, not covered
- Sector rules: health, not covered
- Sector rules: finance and insurance, not covered
- Sector rules: employment, not covered
- Sector rules: education, not covered
The full list of controls could not be loaded, so only the first 10 are shown.
3.3 1 control
- Applies to: AnyoneDrafted by the Index
GenAI providers and deployers should use generative AI to process and analyze multiformat data and simulate potential outcomes to support strategic decision-making in crisis management, healthcare and education.
3.4 2 controls
- Applies to: AnyoneDrafted by the Index
GenAI providers and deployers should generate content adapted for individuals with disabilities to ensure universal accessibility across platforms and communications.
- Applies to: AnyoneDrafted by the Index
GenAI providers and deployers should produce materials in local dialects and aligned with cultural traditions to promote clear communication across diverse populations.
3.5 1 control
- Applies to: AnyoneDrafted by the Index
GenAI providers and deployers should customize public service delivery and expand availability to enhance citizen experiences and satisfaction across healthcare, education, social welfare and legal assistance.
4 6 controls
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should train genAI systems on data that is purged of bias and represents affected minority groups.
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should test genAI models thoroughly to confirm that bias has not been incorporated into code or algorithms.
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should understand the origin, contents, selection and preparation of training data.
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should build knowledge about bias, diversity, inclusion, anti-racism and ethics to recognize discriminatory content.
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should design genAI systems that withstand uncertainty, instability and volatility.
- ObligationApplies to: AnyoneDrafted by the Index
Developers and users should establish standards and protocols for assessing reliability to secure the safety of algorithms and data output.
Change history
No changes recorded since we began monitoring this policy.