Compliance12 min read

UAE PDPL and AI Chatbots: What the Data Protection Law Requires

What Federal Decree-Law No. 45 of 2021 means for a customer-facing AI chatbot in the UAE, and how the DIFC and ADGM regimes differ, written by practitioners rather than lawyers.

Scale model of a records room, a white cabinet closed by a coral padlock and a figure holding a folder: personal data protection for AI chatbots under the UAE PDPL

A customer sends your WhatsApp number their name, budget and Emirates ID photo, and an AI agent answers. Under the UAE's federal data protection law, that agent is processing personal data on your behalf, so the rules on consent, purpose, individuals' rights, breaches and cross-border transfers apply to it exactly as they apply to a human team. If your company sits in the DIFC or ADGM, a different law applies, and the DIFC has rules written specifically for AI systems.

This guide sets out what that means in practice for a business deploying a customer-facing chatbot or AI agent. It is written by practitioners who build these systems, not by lawyers. It is not legal advice. Use it to prepare a better conversation with your counsel, not to replace one.

First, work out which law applies to you

The UAE does not have a single data protection regime.

  • Federal Decree-Law No. 45 of 2021 on the Protection of Personal Data (the PDPL) is the federal law. According to the UAE government portal, it came into force on 2 January 2022 and applies to the processing of personal data by electronic systems, whether in full or in part. The UAE Data Office, set up by a separate decree-law (No. 44 of 2021), is the federal regulator.
  • The DIFC has its own Data Protection Law, DIFC Law No. 5 of 2020, with its own regulations and Commissioner.
  • ADGM has its own Data Protection Regulations 2021, enacted on 11 February 2021 and supervised by the ADGM Office of Data Protection.

Law firm commentary on the PDPL, such as Clyde & Co's overview, notes that it excludes companies in free zones that have their own data protection legislation (DIFC and ADGM are the obvious examples). It also excludes certain categories, including government data, health data governed by other legislation, and banking and credit data governed by other legislation.

So the first question for any chatbot project is simple: where is the company that deploys it established? A mainland Dubai brokerage and a DIFC wealth manager running the same bot face different rulebooks.

Scale model of a records room, a white cabinet closed by a coral padlock and a figure holding a folder: personal data protection for AI chatbots under the UAE PDPL

Under the PDPL, consent is the main basis for processing personal data. The u.ae portal summarises it plainly: the law prohibits processing personal data without the owner's consent, with exceptions.

Law firm summaries describe the consent standard as close to the GDPR's. Consent should be a specific, clear and unambiguous indication of the person's wishes, given by a statement or a clear affirmative action. You must be able to prove it, the request must be clear and easy to understand, and people must be told they can withdraw it at any time.

There are other bases, including where processing is necessary to perform a contract with the person or to comply with a legal obligation. One difference from European law matters for chatbots: the PDPL does not include "legitimate interests" as a basis. If your team is used to GDPR thinking, do not assume you can rely on it.

What this means in a chat:

  • The first message from the agent should say who the person is talking to (your company), that it is an automated assistant, and link to a privacy notice.
  • If you plan to reuse the conversation for marketing, ask for that separately. Answering an enquiry and sending promotions later are different purposes.
  • Log the consent with the conversation, so you can show it later.

Purpose limitation and the training question

Collect what the conversation needs, and use it for what you told the person. A lead qualification bot needs a name, a contact, a budget range and a requirement. It rarely needs a passport scan.

Two practical checks:

  • Do not ask for identity documents in the bot unless the process requires them, and if it does, route them to a secure store rather than leaving them in chat logs.
  • Check your model provider's data terms. Read the terms for the API account you actually use, not the consumer app: whether inputs may be used for training, how long they are kept and where. Record the answers in your processing records.

Individuals' rights inside a chat channel

The PDPL gives individuals rights that include access to their data, correction, erasure, restriction of processing and objection. Clyde & Co's summary also notes a right to object to automated processing, which is directly relevant when a bot scores or filters leads without a human.

Design for these from the start:

  • A person who types "delete my data" or "speak to a human" should get a clear path, not a loop.
  • Someone on your team should be able to find every message and record linked to a phone number, export it and delete it.
  • If the agent makes decisions with real effect, such as rejecting an applicant or ranking a lead, keep a human able to review and overturn them.

Where the data goes: cross-border transfers

Most AI agents send conversation text to a language model hosted outside the UAE. That is a cross-border transfer of personal data.

Under the PDPL, transfers are permitted to countries with an adequate level of protection. Without that, law firm summaries list other routes: a contract that binds the recipient to PDPL-equivalent protections, the person's express consent (where it does not conflict with UAE public and security interests), or where the transfer is necessary for a contract with the person. Much of the detail was left to executive regulations.

Practical steps:

  • Map every service the conversation passes through: the WhatsApp platform, the automation tool, the model API, the database, the CRM.
  • Note where each one stores and processes data, and keep the contracts and data processing terms on file.
  • Where a provider offers regional hosting or data residency, consider whether you need it.
  • Strip what the model does not need. If the bot only needs to classify an enquiry, it may not need the phone number in the prompt.

Breaches: who you tell

Commentary on the PDPL states that a breach likely to put the privacy, confidentiality or security of personal data at risk must be notified to the UAE Data Office and to affected individuals, with timing to be set by executive regulations.

For a chatbot, the realistic breach scenarios are dull ones: an API key leaked in a shared document, a database left publicly readable, an automation sending one customer's details to another. Keep keys in a secrets manager, restrict database access, and test the workflows that send data outward. Write down now who decides whether an incident is notifiable and who contacts the regulator.

DIFC and ADGM: separate rules

DIFC and Regulation 10

The DIFC went further than the federal law on AI. Its Data Protection Regulations include Regulation 10, on personal data processed through autonomous and semi-autonomous systems. The DIFC's own FAQ on Regulation 10 explains the main points:

  • It covers systems that process personal data for purposes defined by humans and/or by the system itself, and that generate an output from that processing. Purely deterministic automation, with no degree of autonomy, is not meant to be captured.
  • It places obligations on the Deployer (broadly comparable to a controller: whoever has authority over the system or benefits from its output) and the Operator (broadly comparable to a processor: whoever runs the system for the Deployer).
  • Deployers or Operators must give users clear and explicit notice when they first use or access the system. The notice should flag processing that is not human initiated, controlled or directed, and explain the impact on users' rights to rectification, erasure or objection.
  • For systems used in high-risk processing, the FAQ refers to an accreditation and certification framework, and describes an Autonomous Systems Officer role that, for now, performs a function similar to a data protection officer.

The DIFC has been consulting on further changes to its regulations in 2026, so check the current consolidated text before relying on any detail.

ADGM

ADGM's Data Protection Regulations 2021 are, in ADGM's own description, closely based on the GDPR. ADGM entities should work from those regulations and the Office of Data Protection's guidance rather than the federal law.

What is still unsettled

Be wary of anyone who tells you the federal picture is complete. The PDPL left a great deal, including breach timing, some transfer mechanics, data protection officer thresholds and penalties, to executive regulations. Older law firm commentary records that the law itself did not set out the penalties. As of November 2026, published sources disagree on whether the executive regulations have been issued in full, and we could not confirm their status against the Official Gazette. Treat that as a question for your lawyer, and build to the law's principles in the meantime: they will not change.

A short checklist before launch

  • Confirm which regime applies: federal PDPL, DIFC or ADGM.
  • Write a privacy notice that covers the bot, and link it in the first message.
  • Record consent, and ask separately for marketing.
  • Collect only what the workflow needs.
  • List every service the data passes through, with its location and terms.
  • Build "talk to a human" and "delete my data" paths.
  • Keep humans able to review decisions with real effect.
  • Set up logging, access control and a breach response owner.
  • In the DIFC, add the Regulation 10 notice and assess whether the processing is high risk.

When we build agents as an AI agency in Dubai, these points go into the first sprint, and every account sits in the client's name so their own team controls the data. If you are still choosing a supplier, our guide to choosing an AI automation agency in Dubai includes the data protection questions to ask.

How we can help

AI agency in Dubai

AI agents, automations and custom apps, shipped in one-week sprints on your own accounts.

See how we do it

Out of Office · Published 10 November 2026

Ready to take action?

A 30-minute call. We look at your stack and your pain points, and tell you what we can ship in 7 days.

Before

Thirteen tools. None of them talk to each other.