Questions people ask

AI (artificial intelligence) is software that does tasks we usually link with human thinking, such as understanding speech, translating, summarising or spotting patterns. Most of the AI people meet today is generative AI: tools that write, answer questions or make images when you ask. It isn't a mind or a person. It learns patterns from very large amounts of data and then predicts a likely response. That is why it can be impressively useful and also, at times, confidently wrong. It is already everyday technology here: Microsoft estimates that 73.3% of working-age people in the UAE used a generative AI tool in June 2026, though other ways of measuring give lower figures [UAE6]. → A calm guide to AI, AGI and ASI

• AI is what exists today: tools that are very good at some tasks and surprisingly weak at others. The International AI Safety Report 2026 calls this "jagged" [K4.1].
• AGI (artificial general intelligence) is a loose term for AI that could handle most intellectual tasks about as well as a capable person. There is no agreed definition or test.
• Superintelligence (ASI) is a hypothetical AI far beyond humans in almost every area. The philosopher Nick Bostrom's widely cited definition is "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest" [K4.5].

AGI and ASI don't exist today. Experts disagree about whether, and when, they might. → Glossary · Full explainer

Superintelligence means AI far more capable than people in almost every field. It doesn't exist, and nobody knows whether or when it will. Views differ widely. Some industry leaders think very powerful AI could arrive within years. Others, such as Turing Award winner Yann LeCun, argue that today's methods won't get there [K4.3, K4.4]. The largest survey of AI researchers gave a combined 50% chance that machines could outperform humans at every task by 2047, and most respondents thought good outcomes were more likely than bad [K4.2]. It is reasonable to take the question seriously without letting it disturb your daily life. The most useful things now, such as human oversight, testing and education, help whatever the timeline. → If AI becomes far more capable than us, how would we live alongside it?

Some respected scientists think it could one day. In 2023 many researchers and AI-company leaders signed a one-sentence statement saying that reducing "the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war" [G5]. Others think this risk is unlikely or far off. In the largest researcher survey, the median estimate for extremely bad outcomes was 5%, and most respondents thought good outcomes more likely [K4.2]. The International AI Safety Report 2026 says current systems "lack the capabilities to pose such risks, but they are improving in relevant areas" [K4.1]. Both things can be true: the risk is taken seriously, and serious work is being done on it. → Fears vs Reality: Could AI threaten humanity itself? · The p(doom) debate

It is a hypothesis: the idea that AI might one day improve itself so fast that the future becomes impossible to predict. The writer and mathematician Vernor Vinge named it in 1993 and expected it by 2030. The futurist Ray Kurzweil popularised it and dates it to 2045 [K4.8]. Vinge's first dated prediction, superhuman intelligence "within thirty years" of 1993, has passed without it happening, and critics argue that real-world limits such as energy, data and experiments would slow any runaway change. Nothing in your daily life needs to change because of it. It is worth understanding because it shapes how some technology leaders think. → The singularity, explained

Many jobs will change. Some tasks will disappear and new ones will appear. So far the evidence is calmer than the headlines. The International AI Safety Report 2026 says: "Early evidence shows no effect on overall employment, but some signs of declining demand for early-career workers in some AI-exposed occupations" [K4.1]. The IMF estimates that about 40% of jobs worldwide are exposed to AI. In advanced economies, roughly half of those may benefit and half may see lower demand [K4.6]. The World Economic Forum expects 170 million jobs to be created and 92 million displaced by 2030 [K4.7]. These are projections, not certainties. The practical step is to learn how AI fits into your own work. → Fears vs Reality: Will AI take my job?

No one can promise which skills will matter most, but some choices are sensible in any scenario. Learn to use AI tools well, and to check their answers. Keep building judgement, communication and subject knowledge, because AI rewards people who can tell a good answer from a bad one. Gulf governments are investing in exactly this. AI has been a formal subject in UAE public schools from kindergarten to Grade 12 since the 2025–26 school year, with about a quarter of the curriculum on ethics [UAE5]. Saudi Arabia introduced an AI curriculum for more than 6 million public-school students in the same year [BIZ4]. Microsoft has committed to help train one million people in the UAE by the end of 2027 (Microsoft's commitment, cited in our engineering article) [UAE19]. → What the Gulf is building: people · AI in everyday life

It can be, with guidance, much as the internet can. Three things help. First, treat AI as a helper, not an authority: children should learn that it can be wrong. Second, watch for over-reliance. The International AI Safety Report 2026 highlights emotional reliance on "AI companions" and people trusting machines too readily as real, everyday risks [K4.1]. Third, protect personal information: children shouldn't share names, photos or locations with AI tools. Ask your child's school how it teaches AI. In the UAE, ethics makes up about a quarter of the new school AI curriculum [UAE5]. Note: we have not yet published a dedicated guide for parents. It is planned. → Fears vs Reality series

It depends on the tool and on what you share. Most GCC states now have personal-data protection laws. Examples are the UAE's Decree-Law 45 of 2021 and Saudi Arabia's law, fully enforced since September 2024 with fines of up to SAR 5 million [K2.2, K2.4, REG3]. The UAE law includes a right to object to decisions based solely on automated processing [REG1]. According to a 2024 legal review, Kuwait has sector rules but no comprehensive law [REG3, REG5]. Laws help, but habits matter most. Don't paste ID numbers, medical records, passwords or confidential work documents into public AI tools. Check the privacy settings, and use your employer's approved tools for work. → Fears vs Reality: Will AI be used to watch us?

AI does make it cheaper to collect and analyse information about people at scale, so this concern is reasonable. The CEO of one leading AI company has written that "AI seems likely to enable much better propaganda and surveillance" [K4.4]. International observers have raised civil-liberty concerns about technology in many countries, including in our region. For most people, though, the most likely everyday privacy harm is fraud, not monitoring: the International AI Safety Report 2026 lists scams and deepfakes among the main current harms [K4.1]. The main protections are data-protection laws, careful system design and accountability. The Gulf now has these laws on the books, and the task is to apply them well. → Fears vs Reality: Will AI be used to watch us? (This article has a paragraph under legal review.)

Trust, but check. AI tools can be excellent for explaining, drafting and summarising. They can also produce "hallucinations": invented facts, quotes or references stated with total confidence. The International AI Safety Report 2026 lists fabricated information and flawed outputs among current problems. It also warns about automation bias, our tendency to believe a machine too readily [K4.1]. A simple rule: the more an answer matters (health, money, law, news), the more you should confirm it with a qualified person or an official source. Ask the tool for its sources, then open them yourself. → Risks, and the conditions they would need · Glossary: hallucination

They are improving, but check them more carefully than English answers. Many AI systems learned mostly from English-language internet data. A 2024 study tested 16 models, including Arabic-focused ones, and found they tended to favour Western names, foods and customs even in Arabic prompts, sometimes with stereotypes [K3.7]. Dialects, religious questions and local laws are areas where mistakes are more likely. The region is building Arabic-first models to close this gap: Jais in the UAE (2023), ALLaM behind HUMAIN Chat in Saudi Arabia (2025) and Fanar in Qatar (2024) [UAE11, KSA8, QA3]. For important matters, confirm with an official Arabic source or a qualified person. → Fears vs Reality: Is AI biased? · What the Gulf is building: models

It can be. AI learns from data, and data over-represents some languages, cultures and viewpoints while under-representing others. If an AI system screens job applications or recommends treatments, hidden assumptions could disadvantage people who don't fit the patterns it learned. That is why testing matters. In the UAE, the Central Bank now expects the financial firms it licenses to test their AI for bias at least once a year and to let customers ask for human review of AI decisions [REG7]. Arabic-first models aim to reduce cultural gaps [QA3, UAE11]. If you think an automated decision treated you unfairly, ask for the reasons and for a human review. → Fears vs Reality: Is AI biased against people like me?

Deepfakes are fake videos, images or voice recordings made to look or sound like a real person, and they are used in fraud [K4.1]. Detection tools aren't reliable enough to depend on, so habits are your best protection. Be suspicious of urgency ("transfer now", "don't tell anyone"). Verify through a second channel you already trust, for example by calling the person back on a number you know. Agree a family "safe word" for emergencies. Never share one-time passwords. For news images, look for the original source. Some cameras and apps now attach "Content Credentials" showing where an image came from, though their authors say this is "not a cure-all" [K5.2]. → Can AI help keep the peace? (verification) · Surveillance and scams

Start small and low-stakes. Try one well-known assistant for something useful but not sensitive: planning a trip, explaining a letter, drafting an email or practising a language. Ask follow-up questions, and ask where it got its information. Keep personal and confidential details out of your prompts (see Q9). Check anything important before you act on it (see Q11). At work, ask what tools and rules your employer has approved: regulated sectors in the Gulf now expect human oversight of AI [REG7]. Above all, stay in charge. AI is a helper you direct, not a decision-maker you obey. → AI in everyday life · Glossary

Several layers of people, none of them perfect on their own. Governments set rules: the UAE Charter for the Development and Use of AI (2024) lists 12 principles including safety, privacy and human oversight [K2.1], and Bahrain adopted a national policy for government use of AI in 2025 [BH1]. Data-protection laws apply across most of the GCC [REG3]. Internationally, more than 100 independent experts produced the International AI Safety Report 2026 [G4]. The same report counts 12 companies that published or updated frontier safety frameworks in 2025, mostly on a voluntary basis [K4.1]. Researchers test systems for failures before they matter [K3.10]. Much of this is new, and whether it is enough is honestly debated. → The safety-focused · Will we lose control of AI?

Governments have set out their reasons in national strategies. The UAE's National AI Strategy 2031 aims to make the country a global leader in AI and projects up to AED 335 billion in extra growth, which is a projection, not an outcome [UAE2, UAE3]. Saudi Arabia's national data and AI strategy aims to place the Kingdom among the top 15 AI nations by 2030 [KSA2]. Qatar, Bahrain, Kuwait and Oman have their own strategies or policies [QA1, BH1, KW1, OM1]. The region also has capital and energy. MGX in Abu Dhabi closed a $49 billion fund [UAE21], and Qatar's Qai and Brookfield announced a $20 billion venture [QA6]. Barakah nuclear plant supplies up to a quarter of the UAE's electricity (ENEC's figure) [K3.2]. Whether the investments pay off is still to be seen. → The GCC AI landscape · What the Gulf is actually building

Both, and the difference matters. Some things are clearly operating: AI is taught in UAE schools [UAE5], Dubai began commercial robotaxi services in March 2026 [K1.6], and Saudi Arabia's Seha Virtual Hospital connects hundreds of hospitals [K1.2]. Other things are still plans. The 5GW UAE–US AI Campus is being redesigned [A4]. As of 28 September 2026 we had found no official confirmation that the first 200MW of Stargate UAE is in service, although its builder said in August that it was on track to hand over capacity in 2026 [UAE17; tracker CR-C01]. A gigawatt announcement is a target, not a working facility. Our public tracker records each major claim, its source, and whether it has been delivered. → Tracker: where are we now? · The GCC AI landscape

A war involving the United States, Israel and Iran began on 28 February 2026, and shipping through the Strait of Hormuz has been heavily restricted since. An interim US–Iran agreement signed on 17 June 2026 broke down in July and expired in August without a final deal. As of late September no agreement was in force [A1]. In March 2026 drones struck AWS data centres in the UAE and a nearby strike damaged a facility in Bahrain. AWS later said some data held only there cannot be restored [A3]. Reuters reported that the UAE–US AI campus is being redesigned as dispersed, better-protected sites [A4]. Much AI work has continued, adapted rather than stopped. Our first thoughts are with everyone affected. → Resilience in 2026 · AI across the Middle East

It uses a lot of electricity, and the amount is growing. The International Energy Agency (IEA) estimates that data centres used about 1.5% of the world's electricity in 2024 and projects this will more than double to around 945 TWh by 2030, slightly more than Japan uses today [G1, K3.1]. The IEA also puts this in proportion: data centres account for about a tenth of global demand growth to 2030, less than air conditioning or electric vehicles. In its words, concerns that AI could accelerate climate change "appear overstated, as do expectations that AI alone will address the issue" [K3.1]. Water for cooling is a real question in a dry region. We haven't yet verified Gulf-specific water figures to our standard, so we don't quote any. → Electricity: the region's quiet advantage

An AI agent doesn't just answer questions. It carries out a sequence of steps towards a goal, such as checking documents, filling in forms or routing a request, ideally with a person approving what matters. In April 2026 the UAE announced a framework to move 50% of federal government sectors, services and operations to agentic AI within two years. It is to be phased in "based on continuous performance and impact assessment", with training for every federal employee [K1.9]. In Sheikh Mohammed bin Rashid's words, "People come first." It is an official target, and progress should be judged as it happens. We track it publicly. → AI in everyday life: dealing with government · Anatomy of an agentic organisation

Intelligence Age Gulf is an independent, bilingual (Arabic and English) knowledge base about AI in the Gulf. It exists to help people here understand AI calmly and accurately, hear every serious point of view fairly, and use AI with confidence. We aren't a promotional outlet and we aren't an alarm bell. The project was started in Dubai. Its legal form, the country where it will be registered, and the names of the founding team and of an advisory panel of academics, safety researchers and Arabic linguists will be published here once they are confirmed.

This is being decided, and we won't guess. The owner is still choosing between funding models: membership fees, sponsorship, events, courses, research, or a mix. What has been set as a principle is independence. Funding and conflicts of interest will be disclosed on a public funding-transparency page. Sponsors will have no say over content or over the status of any claim in our tracker, and board seats will never be for sale. When funding is agreed, it will be published here with names and amounts.

Membership is still a concept, and prices and details haven't been decided. The current idea is a free Community tier for everyone (read the knowledge base and tracker, take part in the forum, and receive a monthly bilingual briefing), plus student, individual, fellow and organisation tiers with extra activities such as events, mentoring and working groups. Forming an association or forum requires licensing in GCC countries, and legal advice is being sought before anything opens. Until then, everything we publish is free to read.

Every factual claim we publish comes from a single, dated fact base with a source link and a verification level. We keep what has been announced separate from what is actually operating. We attribute superlatives ("world's first") to whoever claimed them, and we clearly label speculation. When facts change, we correct them publicly with a dated note. For example, on 28 September 2026 we corrected our description of the June 2026 US–Iran agreement after checking it against Reuters, the BBC and Al Jazeera [A1]. We are setting up a public corrections form. → Tracker: how claims are updated