Saudi and UAE Data Protection Laws for AI Assistants: A 2026 Guide
What Saudi Arabia's PDPL and the UAE's data protection law require when you run chatbots and LLMs: transfers abroad, consent, DIFC Regulation 10, and architecture.

When the AI assistant on a Saudi retailer's website sends a customer's message to a language model hosted in the United States or Europe, that is not just a technical step. If the message contains a name, a mobile number, or an order number linked to a person, it is a transfer of personal data outside the Kingdom, and the Personal Data Protection Law (PDPL) sets specific conditions for it. In the UAE, the federal law has been in force since 2022, but its Executive Regulations have still not been issued, while the Dubai International Financial Centre has binding rules written specifically for AI.
This guide explains what the laws in both countries require of AI assistants and workflows built on large language models, and how to design a project so that compliance does not block it. It is current as of October 2026.
The short answer
- Saudi Arabia's PDPL has been fully enforceable since 14 September 2024, and it is enforced. The violation committees of the Saudi Data and AI Authority (SDAIA) issued 48 penalty decisions in 2025.
- Sending personal data to a language model abroad is a transfer outside the Kingdom. Because no list of adequate countries has been published, such a transfer usually needs a risk assessment and an approved safeguard such as standard contractual clauses.
- The UAE's federal PDPL has been in force since 2 January 2022, but its Executive Regulations had not been issued by October 2026. There is no adequacy list, no model clauses, and no fine schedule yet.
- The DIFC is the only place in either country with binding AI-specific rules, through Regulation 10: a notice at first use, a register of system use, and extra conditions for high-risk processing.
- Health data in the UAE must stay in the country, apart from listed exceptions. In Saudi Arabia, draft amendments open for consultation until 5 November 2026 would make storage inside the Kingdom the default.
- The safest architecture: keep personal data out of prompts where you can, store logs in the country, host the model locally when the data is sensitive, and document your assessments.
Why this matters for AI assistants in particular
A customer conversation is full of personal data: a name, a mobile number, an email address, a national ID or residency number, a delivery address, order details, and sometimes health or financial information the customer volunteers. Every call to a language model that includes this data processes it, and if the model is hosted abroad, it transfers it abroad. Then come the conversation logs the provider keeps, backups, and monitoring tools, each of them further processing.
It becomes more sensitive when the system makes decisions on its own, such as pre-approving a financing request or rejecting a claim, or when it handles special categories of data. An assistant that is merely "smart" is not enough: you need to know where every message goes, why, and on what legal basis.
Saudi Arabia: the Personal Data Protection Law
The PDPL was issued by Royal Decree M/19 of 9/2/1443H (16 September 2021) and amended by Royal Decree M/148 of 5/9/1444H (27 March 2023). It came into force on 14 September 2023 and became fully enforceable after a one-year transition on 14 September 2024. SDAIA supervises it through its National Data Management Office.
The law reaches beyond the Kingdom: it covers the processing of Saudi residents' data by entities outside it. A company in Dubai or Istanbul running an assistant for customers in Riyadh is therefore concerned.
| Topic | What the law requires |
|---|---|
| Legal basis | Consent is the default, with cases where it is not needed, including performing an agreement to which the data subject is a party, and legitimate interest, added in 2023 |
| Legitimate interest | Not available for sensitive data or to public entities, and requires a documented assessment |
| Sensitive data | Racial or ethnic origin; religious, intellectual, or political belief; security and criminal data; biometric or genetic data used for identification; health data; data indicating that one or both parents are unknown |
| Fully automated decisions | Require the data subject's explicit consent |
| Impact assessment | Required in cases including sensitive data, emerging technologies, automated decisions, and continuous monitoring |
| Registration | In the national register of controllers on the National Data Governance Platform, for public entities, controllers whose main activity is processing personal data, and controllers processing sensitive data |
| Data protection officer | For public entities with large-scale processing, and controllers whose core activity is regular monitoring of individuals or processing sensitive data |
| Breach notification | To SDAIA within 72 hours if the breach could cause harm, and to affected people without undue delay |
| Data subject requests | Answered within 30 days, extendable by another 30 |
| Penalties | A warning or a fine of up to SAR 5 million, doubled for repeat offences; up to two years in prison and/or SAR 3 million for disclosing sensitive data with intent to harm or for personal gain |
Credit data was removed from the sensitive list in the 2023 amendments, but it still requires explicit consent.
Related: Multilingual AI Assistants: The Complete Guide to Arabic, Turkish, and English
Sending data to a language model outside the Kingdom
This is the question most AI projects in Saudi Arabia face, because the most capable models usually run on servers outside the Kingdom. The answer comes from the PDPL and the Regulation on Personal Data Transfer outside the Kingdom, first issued by SDAIA Decision 1517 in September 2023 and replaced by Decision 1840, published in the Umm Al-Qura gazette on 1 September 2024.
- Three general conditions (Article 29): the transfer must not harm national security or the Kingdom's vital interests, the data must be adequately protected abroad, and it must be limited to the minimum necessary.
- No list of adequate countries has been published. In practice, transfers therefore rely on the exceptions: limited or one-off transfers, central operations within a corporate group, direct service to the data subject, or scientific research.
- Plus an approved safeguard: standard contractual clauses, binding common rules, or an accreditation certificate. Binding common rules do not cover transfers to vendors outside the group, so an external vendor usually needs the standard contractual clauses SDAIA published in September 2024.
- The "direct service to the data subject" case is narrower than it looks: the recipient itself must hold a Saudi accreditation certificate, and sensitive data is excluded.
- A risk assessment is mandatory before relying on any of these exceptions, and for continuous or large-scale transfers of sensitive data. SDAIA published guidance on transfer risk assessments in February 2025.
What does this mean for an assistant calling a model abroad? In the typical case: standard contractual clauses with the model provider, a documented risk assessment, personal identifiers stripped from messages before they are sent, a check that no sensitive data is included, and all of it recorded in your records of processing. If the data is sensitive or the assessment comes out negative, the answer is a model hosted inside the Kingdom.
Watch what is coming. The draft second round of amendments to the Implementing Regulations, open for consultation on the Istitlaa platform until 5 November 2026, would require controllers to store personal data inside the Kingdom by default, while transfers would remain possible under the PDPL. It would also drop the harm threshold for 72-hour breach notification and require registration for any transfer abroad. It is a draft and has not been adopted, but designing your project now so that logs are stored inside the Kingdom and only the minimum leaves it will save you a rebuild later.
Related: Self-Hosted LLMs for Enterprises: When Local Models Beat the Frontier
The UAE: the federal Personal Data Protection Law
Federal Decree-Law No. 45 of 2021 on the Protection of Personal Data was issued on 20 September 2021 and came into force on 2 January 2022. Its Executive Regulations, which were due within six months, had still not been issued by October 2026. Companies will have six months to comply once they are issued, which the Cabinet can extend.
The practical result: there is no list of adequate countries, no model clauses, and no fine schedule yet, because fines depend on a Cabinet decision that has not appeared. The regulator, the UAE Data Office, never became fully operational, and on 14 June 2026 it was merged into the new Federal Authority for AI and Data.
- Scope: the law covers data of people in the UAE, wherever the controller is. It excludes government data, security and judicial bodies, personal use, health and banking or credit data governed by their own legislation, and entities in free zones with their own data protection laws, such as the DIFC and ADGM.
- Legal basis: consent is the main basis. There is no GDPR-style legitimate interest.
- Transfers abroad: Article 22 allows transfers to countries with adequate protection. Article 23 allows transfers in other cases, including a contract binding the recipient to the law's requirements where the destination has no data protection law, the data subject's express consent, and performance of a contract with them. There is no general standard contractual clause route.
- Breach notification: breaches must be reported "immediately".
- Data protection officer: required for high-risk processing or processing with new technologies, and for large-scale processing of sensitive data. The officer may be based outside the UAE.
- Automated decisions: data subjects can object to decisions made by automated processing.
Because the Executive Regulations are missing, most companies currently rely on clear contractual commitments from the model provider and, where it fits, express consent, and document why they chose that route. Record those decisions now, because you will need to revisit them once the regulations are issued.
Health data: Federal Law No. 2 of 2019 on the use of information and communications technology in health fields requires health data to be stored and processed inside the UAE, except in the ten cases set out in Ministerial Resolution No. 51 of 2021. An AI assistant for a clinic or hospital that handles patient data therefore usually needs a model and storage inside the UAE.
Child safety: Federal Decree-Law No. 26 of 2025 on Child Digital Safety has been in force since 1 January 2026, with compliance due by 1 January 2027. If your assistant may interact with children, check its requirements.
The DIFC and ADGM
Companies registered in these two financial free zones have their own laws, which in some respects are more mature than the federal framework.
Dubai International Financial Centre (DIFC): Data Protection Law No. 5 of 2020 has applied since 1 July 2020 and was amended by Law No. 1 of 2025, in force since 15 July 2025, which lets individuals bring claims directly. Most relevant for AI is Regulation 10 on processing personal data through autonomous and semi-autonomous systems, in force since 1 September 2023:
- A notice at initial use or access to any application or website service that uses such systems. It must explain that some processing is not initiated or directed by a human, whether the system is limited to human-defined purposes, what its outputs are and how they are used, the safeguards built into its design, and any codes or certifications it relies on.
- A register of system use cases and processing activities, kept by the deployer and the operator and provided on request.
- High-risk processing is prohibited unless conditions are met, such as appointing an Autonomous Systems Officer, meeting audit and certification requirements set by the Commissioner, or limiting processing to human-defined or human-approved purposes.
- A certification regime in development: the Commissioner publishes a certification framework and a list of accredited certification bodies, with stricter requirements planned for high-risk processing.
Under the regulation, the "deployer" is the person under whose authority or for whose benefit the system operates, treated as the controller, and the "operator" is the provider that runs the system on the deployer's direction, treated as the processor.
Abu Dhabi Global Market (ADGM): the Data Protection Regulations 2021 require breach notification within 72 hours, and fines can reach USD 28 million.
Note that transfers between the UAE mainland and these free zones can count as cross-border transfers.
Related: Enterprise LLM Security Hardening: Defending AI Workflows Against Prompt Injections and Data Leakage
Saudi Arabia and the UAE compared
| Topic | Saudi Arabia | UAE (federal law) |
|---|---|---|
| Law | Personal Data Protection Law (2021, amended 2023) | Federal Decree-Law No. 45 of 2021 |
| Status | Fully enforceable since 14 September 2024 | In force since 2 January 2022; Executive Regulations not issued |
| Regulator | SDAIA | Federal Authority for AI and Data, after absorbing the UAE Data Office; supervision arrangements still settling |
| Legitimate interest | Available since 2023, with limits | Not available |
| Transfers abroad | Article 29 and the Transfer Regulation; no adequacy list; safeguards plus a risk assessment | Articles 22 and 23; no adequacy list or model clauses yet |
| Breach notification | Within 72 hours if harm is possible | Immediately |
| Fines | Up to SAR 5 million, doubled for repeat offences | No fine schedule issued yet |
| AI-specific rules | Non-binding principles, guidelines, and a risk framework | A non-binding charter; Regulation 10 binding inside the DIFC |
Where the AI rules themselves stand
As of October 2026, neither Saudi Arabia nor the UAE has a national or federal law dedicated to AI. Regulators have, however, published frameworks that show what they expect:
- Saudi Arabia: the AI Ethics Principles (version 2.0, September 2023) with seven principles and four risk levels; generative AI guidelines for government entities and for the public (January 2024); and the National AI Risk Management Framework (2026), which explicitly covers large language models. All are non-binding. The National Data Governance Platform also offers a voluntary accreditation for AI service providers and an AI ethics self-assessment tool. The draft Global AI Hub Law published by the Communications, Space and Technology Commission in April 2025 concerns hosting "data embassies", not regulating AI.
- The UAE: the UAE Charter for the Development and Use of Artificial Intelligence (2024) sets out twelve principles, including data privacy and human oversight, and is non-binding.
These frameworks are not binding, but they reveal what regulators and large customers will ask about: transparency, human oversight, and risk management. And if your assistant also serves people in the EU, the EU AI Act may apply to you.
Related: Does the EU AI Act Apply to Turkish Companies? A 2026 Guide
A practical architecture for a compliant AI assistant
- Strip identifiers before sending. Replace names, mobile numbers, ID and residency numbers, and email addresses with temporary tokens before the message reaches the model, and restore them in the reply inside your own systems. Most customer questions can be answered without knowing who is asking.
- Store in the country, send the minimum. Keep conversation logs and customer data in a cloud region inside Saudi Arabia or the UAE, and send a model abroad only the text it needs to answer.
- Host the model locally for sensitive data. Health, biometric, criminal, and belief data needs a model in the country or on your own servers, especially in the UAE, where health data is restricted to local storage.
- Negotiate the provider's terms. No training on your data, a defined retention period, a known hosting region, and a list of sub-processors, with the Saudi standard contractual clauses or equivalent contractual commitments in the UAE.
- Disclose in Arabic, at the first interaction. Tell users they are talking to an AI and how their data is used, collect the consents you need, and in Saudi Arabia get explicit consent for fully automated decisions.
- Run an impact assessment. It is mandatory in Saudi Arabia for sensitive data, emerging technologies, or automated decisions, and useful in every case.
- Be ready to notify. Log every model call and every access to data, because a 72-hour deadline in Saudi Arabia and "immediately" in the UAE leave no time to go looking for logs after an incident.
Related: AI Acceptable Use Policy: A Free Template for Companies
Common mistakes
- "The UAE Executive Regulations are out, with fines of up to AED 5 million." No official source confirms that the regulations or a fine schedule had been issued by October 2026.
- "The UAE Data Office enforces the law." The office was merged into the Federal Authority for AI and Data in June 2026, and supervision arrangements are still taking shape.
- "Saudi law forbids data from leaving the Kingdom." Not today, outside sector-specific rules. The October 2026 draft, however, would make in-Kingdom storage the default.
- "Standard contractual clauses are enough for any vendor." A risk assessment is also required, and the direct-service case has extra conditions.
- "Every Saudi breach must be reported within 72 hours." Currently only when harm is possible; the draft would remove that threshold.
- "The National Data Management Office replaced SDAIA." The office is part of SDAIA, and SDAIA is the competent authority.
Frequently asked questions
Can we use ChatGPT or other foreign models with customer data in Saudi Arabia?
Yes, with conditions. Sending personal data to a model abroad is a transfer that must meet Article 29 and the Transfer Regulation: a risk assessment, an approved safeguard such as standard contractual clauses, and only the minimum data necessary. It is often easier to strip personal identifiers before sending, so that the text carries no personal data at all. Free personal accounts are not suitable for customer data.
Does data have to be stored in Saudi Arabia?
Not under the general law today, although sector-specific rules may require it. The draft amendments to the Implementing Regulations, open until 5 November 2026, would make in-Kingdom storage the default, so it is wise to design new projects on that basis.
Have the UAE PDPL Executive Regulations been issued?
No, not as of October 2026. There is therefore no list of adequate countries and no fine schedule yet. Companies will get six months to comply once the regulations are issued.
What is DIFC Regulation 10?
It is a binding regulation for entities registered in the DIFC, in force since 1 September 2023, covering the processing of personal data through AI systems. It requires a clear notice at first use, a register of system use, and additional conditions for high-risk processing, such as appointing an Autonomous Systems Officer or meeting certification requirements.
Is there an AI law in Saudi Arabia or the UAE?
Neither country had a comprehensive AI law as of October 2026. Data protection laws apply, along with non-binding principles and guidelines, and Regulation 10 inside the DIFC.
What are the fines in Saudi Arabia?
Up to SAR 5 million per violation, doubled for repeat offences. Disclosing or publishing sensitive data in breach of the law with intent to harm or for personal gain can be punished with up to two years in prison and/or a fine of up to SAR 3 million.
Next steps
An AI assistant serving customers in Saudi Arabia and the UAE needs more than good Arabic: it needs an architecture that knows where every message goes. Orbitra helps you design Arabic-first assistants with identifiers stripped before sending, in-country storage, and locally hosted models where the data requires them, through AI automation consulting. We also train your teams to handle data safely through corporate AI training.
This article is general information and not legal advice. Consult a lawyer licensed in Saudi Arabia or the UAE about your specific situation.
