Insights · AI and data protection

AI at work in the DIFC: Regulation 10, Copilot and ChatGPT explained for IT

If your DIFC firm uses Copilot, ChatGPT or any AI tool that touches personal data, Regulation 10 of the DIFC Data Protection Regulations applies. Here’s what it asks for, and the IT controls that make AI use safe and provable. See also our cyber security services.
The short answer

Regulation 10 of the DIFC Data Protection Regulations, in force since 1 September 2023, covers processing personal data through autonomous and semi-autonomous systems, which in practice means AI. If you deploy one, you must tell people about it when they first use it, use it only for purposes that humans define or approve, design it to be ethical, fair, transparent, secure and accountable, and be able to produce a register of how it is used. High-risk processing also needs certification and an Autonomous Systems Officer. For most firms the practical work is IT governance: controlling which AI tools staff use and what data those tools can reach.

Key points
  • Regulation 10 has been in force since 1 September 2023.
  • The firm an AI system runs for (the deployer) is treated as a controller; a provider operating it (the operator) as a processor.
  • People must get a clear notice when they first use or interact with the system.
  • You must be able to produce a register of use cases, lawful basis, recipients and transfers on request.
  • High-risk processing needs Commissioner certification and an Autonomous Systems Officer, a role similar to a Data Protection Officer.

What does Regulation 10 cover?

The DIFC updated its Data Protection Regulations on 1 September 2023, adding Regulation 10 on processing personal data through autonomous and semi-autonomous systems. It was among the first rules in the region to deal directly with AI.

It uses two key roles. The deployer is the organisation under whose authority, or for whose benefit, the system runs, and it is treated as a controller. The operator is a provider that runs or supervises the system for the deployer, and it is treated as a processor.

In practice, if your DIFC firm rolls out an AI assistant to staff or uses AI on client data, your firm is usually the deployer, and the AI vendor is usually the operator.

What Regulation 10 requires

RequirementWhere it sitsWhat it means day to day
A clear notice at first use10.2.2(a) and (b)Tell people when AI processes their data: the purposes, the limits, how outputs are used and the safeguards in place
Purposes set by humans10.3.1A written list of approved AI use cases, signed off by a person
Ethical, fair, transparent, secure and accountable design10.3.1 and 10.3.2Testing for bias, being able to explain outputs in plain language, security controls and regular reviews
A register, on request10.2.2(g)Use cases, lawful basis, recipients, sub-processors, locations and transfer safeguards, kept up to date
Supporting evidence10.2.2(c) to (f)Risk and impact assessments, and breach assessments, kept ready to show
High-risk processing10.3.3Only with Commissioner certification and an appointed Autonomous Systems Officer

The DIFC also publishes a Regulation 10 Accelerator Framework, a certification assessment tool and a step-by-step portal guide to help firms work through it.

The real risk: client data pasted into public AI tools

For most firms the immediate problem isn’t a sophisticated AI system. It’s a member of staff pasting a client email, a contract or a spreadsheet into a free AI website to save twenty minutes.

That data may be stored outside the DIFC, kept by the provider and outside any contract you control. You can’t evidence it, you can’t include it in your register, and you may not know it happened. Under Regulation 10 and the DIFC’s transfer rules, that is exactly the kind of processing you need to be able to account for.

Banning AI rarely works; people just use their phones. The better route is to give staff an approved tool and put guard rails around it.

IT controls for safe AI use in a DIFC firm

  1. Approve the tools. Keep a short list of approved AI tools, and use web filtering on your firewall to block or warn on the rest.
  2. Use business accounts, not personal ones. Business versions, such as Microsoft 365 Copilot inside your own tenant, keep prompts and data under your contract and controls.
  3. Fix permissions before switching Copilot on. Copilot can surface anything a user can already open. If SharePoint and Teams are overshared, AI makes that visible very quickly.
  4. Label and protect sensitive data. Sensitivity labels and data loss prevention rules can stop confidential client data being pasted into unapproved sites.
  5. Turn on logging. Microsoft Purview audit can record Copilot activity, giving you evidence of how AI is used.
  6. Keep the register. Maintain the AI register with your DPO, and update it whenever a new use case is approved.
  7. Write a short AI policy and train staff on what they can and can’t put into AI tools.
  8. Check data location for every AI tool, and record a transfer assessment where data leaves the DIFC.

Do you need ISO 42001?

No, Regulation 10 doesn’t require it. ISO/IEC 42001 is the international standard for an AI management system, and it gives you a structured way to govern AI: policies, risk assessment, roles and continual improvement.

If you already hold ISO 27001, you’ll find much of the structure familiar, and it can be a sensible framework to follow even without certification. For most small and mid-sized DIFC firms, getting the basics in this guide right comes first.

Is anything changing?

The Commissioner consulted in 2026 on further changes to the Data Protection Regulations, including how accreditation and certification schemes are recognised. At the time of writing these were proposals, not law. We’ll update this guide when anything changes, and our guide to the DIFC data protection law covers the wider rules AI sits within.

The bit where we mention ourselves

Cre8 IT helps DIFC firms adopt AI without losing control of their data. Before a Copilot roll-out we review permissions, sensitivity labels, data loss prevention and logging, and we set up the web filtering and policies that keep staff on approved tools. We’re ISO 27001:2022 certified and we work alongside your DPO; we don’t give legal advice.

Planning to switch on Copilot or another AI tool? Book a free review first.

Sources and further reading

Last reviewed 9 November 2026. This guide explains the IT side of the law and is not legal advice. Check the current position with the regulator or your legal adviser.

Written by
CEO, Cre8 IT
Rob Crossley is CEO of Cre8 IT, a Dubai-born managed IT and cyber security company that has supported businesses across the UAE, Saudi Arabia and the UK since 2012. Cre8 IT is certified to ISO 27001:2022 and ISO 9001 and is a certified Microsoft Partner.
FAQ

Frequently asked questions

Does Regulation 10 apply to Microsoft Copilot or ChatGPT?

If a DIFC firm uses an AI tool to process personal data, Regulation 10 will usually apply, with the firm as the deployer. That includes AI assistants such as Copilot or ChatGPT when staff use them on client or employee data.

When did DIFC Regulation 10 come into force?

On 1 September 2023, as part of the updated DIFC Data Protection Regulations.

What is an Autonomous Systems Officer?

A role that deployers and operators must appoint before using AI for high-risk processing. The officer has the same or substantially similar competencies, status and tasks as a Data Protection Officer.

Do we need certification to use AI in the DIFC?

High-risk processing through AI is only allowed where the Commissioner has set audit or certification requirements, the system meets them, purposes are set by humans and an Autonomous Systems Officer is appointed. For other uses you still need the notice, the register and the design principles. The Commissioner publishes a certification assessment tool to help you check.

Related guides

Keep reading

Each guide links to the regulator’s own text and shows when it was last reviewed.
DIFC data protection

DIFC data protection law: what it means for your IT, and a 2026 checklist

DIFC data protection law (DIFC Law No. 5 of 2020) explained for IT: the 2025 amendments, breach reporting, DPO rules, transfers and a 12-point checklist.
7 min read · Reviewed 9 November 2026
UAE data protection

UAE data protection law (PDPL): what it means for your IT, and which law applies to you

UAE data protection law explained for IT: PDPL vs DIFC and ADGM rules, breach reporting, DPOs, security, transfers and what to do while regulations are pending.
5 min read · Reviewed 9 November 2026
Microsoft 365

Microsoft 365 governance for regulated firms: a practical baseline

How to govern Microsoft 365 in a DFSA- or FSRA-regulated firm: identity, devices, data, logging and licences, with a 30-day order of work.
4 min read · Reviewed 19 September 2026

Planning to roll out Copilot?

We’ll check your permissions, labels and logging first, so AI only sees what it should and you can prove it.