The EU AI Act applies to companies in Dubai, Abu Dhabi, Riyadh and Doha more often than their boards assume. This page is my EU AI Act UAE and GCC guide: the three ways a Gulf company falls in scope, four worked examples, the role chain, the timeline after the Digital Omnibus and a twelve-month checklist. State of play as of November 2026, to my knowledge. I am an engineer, not a lawyer: this is no legal advice, verify it with counsel.
I have spent much of the last two years in the Gulf: our SIFAMO AI Summit in the Burj Khalifa in February 2025, covered by Gulf News, and two weeks of meetings in Dubai, Doha and Abu Dhabi in summer 2026 with our CTO Ayman Kalaie. One question came up in almost every conversation: we serve European customers, does this law apply to us? More often than you would like.
EU AI Act UAE: three ways a Gulf company falls in scope
The AI Act, Regulation (EU) 2024/1689, does not ask where your company is registered. It asks what you do with an AI system and where the effects land. Three routes in Article 2 matter for the GCC.
- You place an AI system or a GPAI model on the EU market (Art. 2(1)(a)). Providers are covered wherever they are established. A SaaS subscription sold from Dubai to a company in Munich is enough.
- The output of your system is used in the Union (Art. 2(1)(c)). Providers and deployers in a third country are covered where the output is used in the EU. The model runs in Doha, the credit decision lands in Europe.
- You have an EU subsidiary or importer. That entity carries its own duties. If it brings the parent’s system onto the market under the parent’s name, it is an importer under Art. 23; a reseller is a distributor under Art. 24.
Out of scope: a system that stays in the Gulf, serves Gulf customers and whose output never reaches the EU. In a region where almost every serious company has European clients, that exception is rarer than the rule.
Four worked examples
A Dubai fintech sells a credit scoring model to an EU bank
The fintech is the provider, the bank the deployer. Creditworthiness assessment of natural persons is high-risk under Annex III, point 5(b). As of November 2026 the fintech therefore needs, before market entry, an authorised representative by written mandate under Art. 22, Annex IV documentation, risk management, logging, human oversight, a conformity assessment and an EU database entry. The bank has deployer duties under Art. 26 and a fundamental rights impact assessment under Art. 27.
A Riyadh HR-tech vendor whose screening tool is used by a German employer
Screening and ranking candidates is Annex III, point 4. Offered to EU employers, the tool is placed on the market under Art. 2(1)(a); run from Riyadh with shortlists sent to Germany, its output is used in the Union under Art. 2(1)(c). Either way the vendor is the provider. The employer is the deployer: it informs applicants, keeps the logs, ensures oversight.
A Doha insurer with EU policyholders
Here the Gulf company is the deployer. If a Doha insurer prices life or health cover for EU policyholders with a model, the output is used in the Union and Art. 2(1)(c) applies. Life and health pricing is Annex III, point 5(c), so Art. 26 deployer duties apply, and a customer-facing chatbot carries Art. 50 transparency duties on top. As of November 2026 a QCB-licensed insurer is also under the Qatar Central Bank’s AI Guidelines.
A UAE GPAI model provider
The UAE builds its own large models. A provider that offers a general-purpose AI model on the EU market is covered under Art. 2(1)(a) wherever it sits. As of November 2026 the GPAI obligations have applied since 2 August 2025: technical documentation, information for downstream providers, a copyright policy, a training-content summary and, under Art. 54, an authorised representative. The systemic-risk tier adds evaluation, incident reporting and cybersecurity duties.
The role chain and Art. 25
The Act distributes duties along a chain. It is the map of who does what, and where a role can change unnoticed.
- Provider. Develops a system or model and places it on the market under its own name. Heaviest duties, from risk management to post-market monitoring.
- Authorised representative (Art. 22 for high-risk systems, Art. 54 for GPAI models). A person or firm in the EU, mandated in writing before the system reaches the market. It holds the documentation; it does not replace your own compliance work.
- Importer (Art. 23). An EU entity that places a system bearing a third-country name on the market and checks that conformity assessment, documentation and representative exist.
- Distributor (Art. 24). Makes the system available without being provider or importer; checks marking and documentation.
- Deployer (Art. 26). Uses the system under its own authority: follows the instructions for use, assigns oversight, keeps logs, informs affected persons.
Art. 25 is the clause Gulf companies miss most often. A deployer, importer or distributor becomes the provider, with all provider duties, if it puts its own name on a high-risk system, substantially modifies it or changes its intended purpose so that it becomes high-risk. Role follows what you do, not what the contract calls you.
EU AI Act UAE timeline: what applies when
As of November 2026 the dates look like this. Prohibitions and AI literacy apply since 2 February 2025, GPAI obligations since 2 August 2025, transparency duties under Art. 50 since 2 August 2026. The Digital Omnibus on AI (Regulation (EU) 2026/1744) moved the high-risk obligations: Annex III systems from 2 December 2027, Annex I products from 2 August 2028.
Three things follow. First, parts of the Act bind you today: a chatbot talking to EU users falls under Art. 50 now, and the AI literacy duty under Art. 4 applies to staff whose work touches the EU. Second, December 2027 is closer than it sounds: a representative, documentation, an evaluation set the business owns and a conformity path take longer than a year. Third, EU buyers do not wait for the legal date; banks and insurers are writing high-risk requirements into vendor contracts in 2026.
EU AI Act UAE checklist: the next twelve months
The sequence I run with Gulf clients. Each step produces an artefact a buyer, an auditor or an authority can read.
- 1. Classify. Inventory every AI system that touches the EU. For each: prohibited, high-risk under Annex I or III, transparency-only under Art. 50, or minimal, and which role you hold.
- 2. Decide on an authorised representative. Third-country providers of high-risk systems or GPAI models choose one, sign the written mandate and hand over the documentation.
- 3. Build the system registry. One record per model or agent: owner, purpose, risk class, scope, version, evaluation results, expiry.
- 4. Write the technical documentation. Annex IV, from intended purpose to risk management. As of November 2026 most harmonised standards are still drafts, so a well-documented solution of your own is the practical path.
- 5. Switch on the evidence log. Art. 12 asks for automatic logging over the system’s lifetime. Append-only, per decision: input reference, model version, sources, policy version, output, reviewer action.
- 6. Choose the conformity path. Art. 43: most Annex III systems can go through internal control, some cases need a notified body.
- 7. Meet Art. 50 now. Chatbots disclose that they are machines, synthetic content is marked, deepfakes are labelled. Already in force.
One set of controls, two rulebooks
European rules do not double the work on top of DIFC Regulation 10 and the PDPL, if you build the controls once and map them twice.
As of November 2026, to my knowledge, no GCC state has a general AI law in force. The UAE applies its federal Personal Data Protection Law (Federal Decree-Law No. 45 of 2021), and in June 2026 the Cabinet approved a federal Artificial Intelligence and Data Authority, details pending. In the DIFC, Regulation 10 under Data Protection Law No. 5 of 2020 has been in force since 1 September 2023, with its own Deployer and Operator roles, enhanced duties for high-risk systems and a certification-based approach. Saudi Arabia has SDAIA, the PDPL in force since September 2023 and a draft Responsible AI Policy in consultation since April 2026. Qatar has binding QCB AI Guidelines for licensed firms and Personal Data Privacy Protection Law No. 13 of 2016.
Read what these texts ask for and the overlap with the AI Act is large: accountability, risk assessment before deployment, data protection, human control, the ability to explain and to stop. A governance layer does exactly that, whichever rulebook is asking: a system registry, a policy gateway in front of every model call that enforces scope and pseudonymises personal data, an evidence log, tiered human oversight, quality gates and a kill switch. The registry answers the DIFC Commissioner and feeds the EU database, the evidence log serves Art. 12 and the PDPL’s accountability principle, the oversight tiers satisfy Art. 14 and Regulation 10’s enhanced duties. Details in my post on the governance layer.
One caution: the DIFC’s Deployer and the AI Act’s deployer are not the same term, and the DIFC Operator has no twin in the Act. Map roles system by system with counsel.
Penalties
Fines run up to 35 million euros or 7 percent of worldwide annual turnover for prohibited practices, and up to 15 million euros or 3 percent for most other breaches. As of November 2026 enforcement against a company outside the EU runs through its authorised representative, its importer and its EU customers. The commercial penalty arrives earlier: a Gulf vendor that cannot answer an EU bank’s AI Act questionnaire loses the tender before any authority writes a letter.
Where I can help
I work on this from Frankfurt with a team that also builds the systems: remotely, plus on-site phases in Dubai, Riyadh or Doha, in English or German. SIFAMO has no office in the UAE. The regional offer is on my AI governance consultant Dubai page, the pan-European view on EU AI Act consultant, and banks and insurers under DORA should read DORA and the EU AI Act together. All my articles on the Act are in the EU AI Act hub; the guide itself is here.
About Umair Zaffar
I am the founder and CEO of sifamo GmbH, an AI consultancy for regulated industries in Berlin and Frankfurt am Main: founded in 2019, self-financed, 28 people today. I come from enterprise architecture and built large platforms for banks, insurers and the public sector for more than ten years before we focused on AI in production in 2023. Our clients include Commerzbank, VW Bank, Bosch, DHL, Zurich, VGH and the German federal administration. In the Gulf we hosted the SIFAMO AI Summit 2025 in the Armani Ballroom of the Burj Khalifa and are in early talks with Visit Qatar about a summit in Doha in 2027. More on my AI expert page and my founder page.
Frequently asked questions
Do we need an authorised representative if only our output is used in the Union?
Art. 22 ties the representative to providers placing a high-risk system on the EU market. A Gulf deployer whose output is used in the Union has deployer duties, not that one. For a provider that never formally places the system on the market but whose output is used there, the reading is, as of November 2026, less settled than I would like. Ask counsel.
Can we wait until December 2027?
For Art. 50, AI literacy and GPAI duties, no, they apply already. For Annex III systems the legal date is December 2027, but EU customers will ask for the evidence in 2026 and 2027, and a representative, documentation and a conformity path are not built in a quarter.
Does a DIFC Regulation 10 certification count towards the AI Act?
Not formally. As of November 2026 there is, to my knowledge, no mutual recognition between the DIFC regime and the AI Act. In substance the evidence overlaps heavily: a registry, a risk assessment, documented oversight and a log serve both.
Want to know whether the EU AI Act applies to your company in the UAE, Saudi Arabia or Qatar, and what to do first? Reach me via the contact page, by e-mail at contact@sifamo.com or on LinkedIn. On stage I cover this as an AI keynote speaker in Dubai and the Gulf.