AI and work in the Gulf

One Gulf state has written the over-reliance problem into a national governance instrument. Saudi Arabia's AI ethics principles ask system designers a question no other document in the region asks: does your design prevent overconfidence in, or over-reliance on, the AI system. That single line is further than most Western AI frameworks have gone, and almost nobody outside the Kingdom has noticed it.

What was checked, and what could not be

Every document below was read at the issuing body's own page on 28 August 2026, or it is marked and excluded. That distinction does more work here than usual, because the Gulf policy layer is badly served by secondary coverage: several of the most-quoted facts about AI in the region trace back to press releases rather than to any government page, and some of them could not be confirmed at all.

Three limits before the findings. SDAIA's main website rejects automated requests, so Saudi Arabia's strategy document, its adoption framework and its leadership pages could not be read; one subdomain served the ethics document and nothing else. The UAE's state news agency serves no text to a fetcher, so the widely reported national school AI curriculum appears nowhere in this account. And Dubai's richest material on this subject is published in Arabic only.

Saudi Arabia: the only instrument that names over-reliance

SDAIA's AI Ethics Principles, version 1.0 of September 2023, carries an assessment checklist covering the system lifecycle. At the plan and design stage it asks:

"Does your AI system design prevent overconfidence in or overreliance on the AI system with necessary human intervention mechanisms?"

Alongside it: whether human oversight processes carry defined KPIs with responsibility assigned to named parties, and whether training has been provided to develop accountability practices. The operative text states that decisions which are irreversible or life-and-death "should trigger human oversight and final determination", and rules out social scoring and mass surveillance.

Three qualifications travel with that quote. The over-reliance language sits in a checklist annexe rather than in the principle text. The instrument is guidance, not statute: Saudi Arabia had no binding AI law as of August 2026, only a draft responsible AI policy put out for consultation on 2 April 2026. And version 1.0 dates from September 2023, so whether the language survives into any later version is unknown, because the host that would answer that question is closed.

The only real AI statistic in the Gulf

Saudi Arabia's General Authority for Statistics reports that 33.1 per cent of establishments use AI technologies, up 20.0 per cent on 2024, with information and communication at 61.1 per cent, financial and insurance at 52.9 per cent and education at 51.0 per cent. The methodology is stated as aligned with UNCTAD standards.

Two things about that number matter more than the number. The companion household survey for the same year contains no AI indicator at all, so the Kingdom measures firms and leaves citizens unmeasured. And neither the UAE nor Qatar publishes any official AI statistic whatever. Every UAE adoption percentage in circulation, including the ones quoted confidently in conference keynotes, is vendor-produced.

Saudi Arabia is also the only Gulf state where training throughput can be tracked against a stated national target. Its SAMAI programme passed one million citizens trained in November 2025, on the Ministry of Education's own figures, of whom 52 per cent were women and 70 per cent already employed. By June 2026 the state news agency reported 1,563,983 beneficiaries including 14,495 specialists and experts, plus 278 trained government leaders. An AI curriculum entered all levels of public education from the 2025-26 academic year, reaching more than six million school students.

Read carefully, that is training throughput rather than employment, and certificates rather than demonstrated capability. It is the distinction this research keeps returning to at measuring adoption properly. But it is a considerably better evidenced position than either neighbour can show.

The United Arab Emirates: a principle without a duty-holder

The UAE Charter of 10 June 2024 states human oversight as Principle 6, emphasising "the irreplaceable value of human judgment and human oversight over AI". There is no named duty-holder, no enforcement mechanism, no competence requirement and no test of whether oversight is real. The charter is silent on deskilling and over-reliance.

The binding instrument sits elsewhere and covers narrower ground. DIFC Regulation 10, enacted September 2023 and updated August 2024, governs personal data processed by autonomous systems inside one free zone. Its reasoning is the most useful thing in the UAE corpus for this research, because it reaches for the same analogy this estate uses: where a system operates for its deployer, "its position is substantially similar to that of an employee within the Deployer organization, and the Deployer should be therefore liable for its actions in the same way it may be liable for an employee's actions". It creates a named Autonomous Systems Officer.

The analogy is about liability rather than skill. It does not require that the responsible human could do the work being supervised, which is the question at who supervises work they cannot do themselves.

Dubai's 2019 ethics guidelines go furthest of anything in the Emirates on competence, requiring that organisations operating AI hold "sufficient knowledge of the nature of the AI systems they use so as to be able to know their suitability for the use case", and that systems informing significant decisions face quality-checking "equivalent to those applied to a human employee taking that type of decision". That is close to this estate's own argument. It is also non-binding, seven years old, published in Arabic only, and its external audit provision is marked suspended for want of any audit mechanism.

On scale of ambition the UAE leads: a May 2026 Cabinet programme to train 80,000 federal employees in agentic AI tools across five occupational categories, and a national target to deploy agentic AI across half of government services. Abu Dhabi reports over 95 per cent of its 30,000-plus employees completing AI training, and frames it explicitly as enhancing rather than replacing human-centred public service. In June 2026 the Office of AI was folded into a new Artificial Intelligence and Data Authority reporting directly to Cabinet.

Qatar: control without competence

Qatar's national AI strategy is a 2019 blueprint produced by the Qatar Computing Research Institute and adopted by the ministry. It carries no named author, no budget, no KPIs and no implementation schedule, and it has never been publicly revised. Its labour analysis rests on the 2016 labour survey. The associated ministerial guidelines carry no publication date, no version number and no reference number, and are not retrievable from the ministry's own website.

On substance, Principle 8 states that "AI systems should not be able to autonomously make decisions of significant consequence" and should "provide users with the ability to appeal or override decisions that have a substantial impact on individuals or society".

Two corrections to how this document is usually described. The phrase human in the loop appears nowhere in it; the operative concepts are human intervention, override and ultimate accountability. And Principle 7, "develop a human-centered approach", is not an oversight principle at all: its items concern cultural and religious values, feedback, diverse teams and accessibility. Anyone citing Qatar's human-centred principle as an oversight duty has misread it.

The gap is clean enough to name. Qatar requires humans to retain control of consequential decisions and imposes no obligation whatever that those humans be trained, assessed or kept current. Control without competence is the arrangement this research argues is the least stable of all, at human in the loop is not a safeguard.

What none of them say

• The word deskilling appears in no verified document from any of the three countries. The concept is absent from the region's AI governance vocabulary.
• Every instrument that touches competence is non-binding guidance. Saudi's ethics principles, the UAE Charter, Dubai's guidelines and Qatar's principles are all voluntary. The only binding text found, DIFC Regulation 10, is a data protection regulation in one free zone.
• None of the three publishes AI employment data. Not a single series on AI-related jobs, in any of them.
• Nobody requires that a human overseer be able to do the work. Every framework assumes the human in the chain is capable, and none asks whether that is still true.

Claims in wide circulation that this page does not make

Named because excluding them silently would be the easier and worse choice.

• The UAE national school AI curriculum. Reported worldwide, including the claim that the UAE is the first country to mandate one across all school years. Not one element could be confirmed at an issuing body's page, because the state news agency serves no text and the education ministry's English media centre does not carry it. It may well be accurate. Verification was not possible here.
• Saudi Arabia's AI investment and specialist targets. The widely quoted twenty thousand specialists and the dollar investment figures could not be confirmed at SDAIA. The only figure traceable to an official-adjacent source is in Saudi riyals, and the frequently cited dollar version has no basis found.
• Qatar's guidelines dated May 2024. The documents carry no date at all.
• Qatar Central Bank's AI guideline. Issued 4 September 2024 and the only binding AI instrument in the country. Its text could not be read, so the human oversight requirements attributed to it by law firms are not repeated here.

What this means for an organisation operating in the region

• Do not assume the framework protects you. Every competence-adjacent provision in the Gulf is voluntary, and the one binding instrument addresses liability rather than skill.
• Saudi's checklist question is a usable internal test. Whether or not you operate there, asking whether a deployment prevents over-reliance is a better design question than asking whether a human reviews the output.
• Training volume is the region's proxy for capability, and a poor one. A million certificates and eighty thousand trained employees measure delivery, not whether anyone can still work unaided.
• Check the date and the source on any Gulf AI fact before you put it in a board pack. A striking proportion of what circulates is press-sourced and some of it is unverifiable.

Key research and primary sources

• Saudi Data and AI Authority (2023). AI Ethics Principles, Version 1.0.
• General Authority for Statistics, Saudi Arabia (2025). Establishments ICT Access and Usage Statistics 2025.
• Government of the United Arab Emirates (2024). The UAE Charter for the Development and Use of Artificial Intelligence.
• DIFC Commissioner of Data Protection (2024). Regulation 10 on Personal Data Processed through Autonomous and Semi-Autonomous Systems.
• Ministry of Communications and Information Technology, Qatar (undated). Artificial Intelligence in Qatar: Principles and Guidelines for Ethical Development and Deployment.

Related SuperSkills research

For the wider international position, AI and work by country and Japan. For the oversight argument these documents circle, human in the loop is not a safeguard and meaningful human oversight. For the measurement problem, measuring adoption properly and usage theatre. For the category, human capability in the age of AI.

About this research

Written by Rahim Hirji, author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. He worked for the government of Abu Dhabi earlier in his career and has travelled and worked across the Gulf since. Every document cited here was read at the issuing body's own page on 28 August 2026; the documents that could not be reached are named rather than paraphrased from secondary coverage.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.