We often hear the elite class telling us AI is “inevitable,” and not so long ago Reese Witherspoon lamented her worries about women being “left behind” by the unstoppable tidal wave of AI heading for us all. 

I am an AI sceptic, with too many ethical concerns to simply accept this vision of the future without resistance. At the same time, though, I want to be honest. My relationship with AI is best told in two stories.

The first occurred not long ago; I was steadfastly against the use of AI and had managed to keep a healthy distance. But then, working to a tight deadline, I made an error on CapCut that undid me. I had spent three hours editing a video project; trimming, layering, getting every cut exactly right, adding sound effects, visual effects–you get the picture. And then, in one careless moment, I deleted the clips from the project. Beads of sweat started to form on my brow as I frantically tried to restore what I had just done. I couldn’t find a way to do so. I started searching the internet, scouring Reddit threads. They all seemed to say CapCut is a merciless mistress, undoing a mistake like mine was not possible!

I am not proud of what happened next: I fell to the floor quite dramatically and moaned. All the while, the clock was ticking, and I was only too aware of my deadline, whilst completely fatigued by the hours of work already burned through. In desperation, I typed the exact problem, step by step, into an AI tool. It offered several possible solutions, walked me through them, and within minutes my video was restored, every edit intact. That was the day my sternness softened. 

The second story is more recent, and it travelled further than I expected when I wrote about it on Instagram. I cancelled an event I was due to speak at because of AI slop. The organisers had produced a promotional poster using AI image generation. The first version rendered both me and my book unrecognisable. I made clear this was unacceptable and asked them to fix it. The second version was worse: the AI was absolutely hallucinating, and the “me” on that poster was now someone else entirely. So, I withdrew. I heard their first reaction was to suggest I was overreacting, but for me, presenting your guest speaker in their actual likeness is the very basics of good manners (adab).

I understand we are overstretched and under-appreciated, and anything that creates ease gets grabbed with both hands. But in a world of free Canva and countless editing apps, there is no excuse for AI-hallucinated images of your speaker. Particularly when you have clearly taken her actual picture from her social media and fed that to the AI tool, that too without her consent. 

Between these two experiences is where I have sat to write this article because AI is a tool, and like all tools, the question is never simply whether we use it, but how, for what, at whose expense, and with what intention. My rescue from CapCut hell and my indignation at the hallucinated poster came from the same technology.

I think what differed was everything our tradition actually cares about: purpose, care, honesty, and respect for the human being in front of you.

My personal experience aside, the question of AI matters because it is no longer at the edges of Muslim life. It is producing halaqa-style videos with synthetic voices reciting reminders over stock footage. It is behind AI-generated “Muslim influencer” profiles; faces that don’t exist, delivering “Islamic” content to real followers – and even collecting charity donations. And it also sits in the quieter rigmarole of life, an almost ever-present undercurrent in daily life: the sister asking a chatbot how to word a difficult message, the student summarising notes, the parent settling a dinner-table debate.

If our communities are using these tools, and we are, then we need an ethical framework.

Refusing the Prophecy of “Inevitability”

Before any framework, though, I want us to go back and dismantle one pesky word in the conversation on AI: inevitable. It is the word behind every slop poster and every claim of “this is just how things are done now, adapt or be left behind.” Oxford philosopher Carissa Véliz takes that word apart in her new book Prophecy: Prediction, Power, and the Fight for the Future, from Ancient Oracles to AI (2026). Her argument is that today’s tech executives play the same role as the oracles of the ancient world: their forecasts are not neutral observations but power plays. Tech leaders have financial incentives to present their preferred futures as already written, and when we believe a prediction unreflectively, Véliz argues, we are closer to obeying an order than assessing a fact. Worse, predictions about the social world tend to be self-fulfilling, so the more we act as though total AI adoption is inevitable, the more we build the world in which it becomes so. 

Muslims, of all people, should recognise a false prophet’s business model. Our theology reserves knowledge of the future for Allah ﷻ alone, “no soul knows what it will earn tomorrow” (Surah Luqman 31:34), and our ethics insists that human choices remain choices, for which we will be held accountable.

“AI is inevitable” is not the qadr of Allah, but the sales pitch of technocrats.

When the organisers of my event treated an AI poster as simply the way things are done now, they were obeying the prophecy written in Silicon Valley, for Silicon Valley’s benefit. The first step of an Islamic AI ethic, then, is simply that we will not be subordinated to technology and we will choose whether, when, how, and for what AI should be used. 

An I’timāni Ethic of AI

There is a growing body of serious Muslim scholarship on AI ethics. Raquib, Channa, Zubair and Qadir (2022) have proposed an Islamic virtue-based ethics for AI grounded in the maqāṣid al-sharī’ah, the objectives of preserving faith, life, intellect, lineage and wealth, arguing that AI cannot be assumed beneficial simply because it is efficient, and that we must interrogate the why of a technology, not just the how. Crucially, they invoke the juristic principle of sadd al-dharā’i, blocking the means to harm, and the classical maxim that warding off harm takes priority over acquiring benefit. That maxim, incidentally, is precisely why cancelling an event over a hallucinated poster is not an overreaction; it is the embodiment of Islamic ethics.

More recently, Ali, Bouzoubaa, Gelli, Hamzi and Khan (2025) have gone further, arguing that legal frameworks alone cannot carry the ethical weight, drawing on the Moroccan philosopher Taha Abdurrahman’s i’timāni (trusteeship) framework. Its foundation is the Qur’anic concept of amānah, the divine trust humanity accepted when the heavens and mountains refused it (Surah al-Ahzab 33:72), unfolding into three covenants of responsibility. These are responsibility towards God, towards knowledge, and towards creation. These three responsibilities in turn map uncannily well onto AI’s three great ethical problems: truth, bias and data, and the earth itself.

Knowledge has a Chain, and AI Doesn’t

The early Muslims guarded knowledge with a rigour the internet can barely imagine. Muhammad ibn Sirin’s words still apply today: “This knowledge is religion, so look carefully from whom you take your religion.”

An AI has no isnad, no teacher, no taqwa, no accountability before Allah, not even a community of peers to check it.

It generates plausible text by pattern, which means it can, and most often does, produce confident, fluent, entirely fabricated answers, including invented hadith, tafsirs, and scholarly quotes, complete with fake attributions. For a community whose Prophet ﷺ warned, “Whoever lies about me deliberately, let him take his seat in the Fire” (Bukhari and Muslim), we cannot brush this off as a minor technical glitch, and I would like to think that most of us would recognise this clearly as a spiritual hazard. But this ethic of integrity and truth must permeate all our curiosities and knowledge-seeking, not just our Islamic knowledge.

This is what Abdurrahman’s epistemological covenant names:

knowledge is itself a trust, and the Qur’an commands, “Do not mix truth with falsehood or conceal the truth while you know it.” (Surah Al-Baqarah 2:42).

The same applies to AI halaqa content and synthetic “Muslim” creator profiles. There is something deeply uncomfortable about deen delivered by a person who does not exist, whose piety is an artificial rendering.

The latter especially worries me when so much of our deen has been reduced to an outward “aesthetic” and little by way of contemplative inner work. The Qur’an commands verification before transmission: “O you who believe, if a wicked person comes to you with news, verify it (tabayyun).” (Surah Al-Hujurat 49:6)

A machine is not wicked, sure, but it is also not truthful, because truthfulness requires intention. Tabayyun means that anything AI tells us about Islam must be checked against scholars, texts, teachers, and of course, your own common sense, before we act on it or repost it as a reminder with an accompanying humming nasheed for added “legitimacy.”

The Biases Baked into the Algorithm

No age, teacher, or field is free from bias, and nor can we pretend AI tools are, because they carry within them the very biases of those who create them. In her 2025 book The New Age of Sexism: How AI and Emerging Technologies Are Reinventing Misogyny, Laura Bates documents how sexism is being coded into the foundations of emerging technology; from AI systems trained on datasets saturated with society’s prejudices, to deepfake pornography that overwhelmingly targets women and girls, to hiring algorithms that learned to downgrade women’s CVs. Her argument is stark:

not only do these technologies fail to fix old inequalities; they are amplifying them at both scale and speed.

For Muslims, this should land twice as hard, because the bias is racialised and Islamophobic too. Ali and colleagues open their paper by citing research demonstrating persistent anti-Muslim bias in large language models. These systems are shown to associate Muslims with violence even after attempts at debiasing. This is hardly surprising when there is substantial evidence that the financial and political interests behind the most popular AI systems lean towards the Trumpian right. Venture capitalists Marc Andreessen and Ben Horowitz, whose firm is a major investor in OpenAI and Musk’s xAI, publicly endorsed Trump in 2024, with donations from these tech founders and their wives too, amounting to hundreds of millions of US dollars in donations to pro-Trump PACs.

Musk in particular has an extensive record of promoting far-right, anti-migrant and anti-Muslim politics, repeatedly endorsing Germany’s AfD, an anti-immigrant, anti-Islam party monitored by the country’s domestic intelligence agency, and amplifying racist disinformation targeting British Muslims of Pakistani heritage. And Musk’s chatbot Grok offers a warning of how directly an owner’s ideology can shape a model’s outputs: after programmers under Musk’s direction worked to strip “woke ideology” from its replies, Grok praised Hitler, engaged in Holocaust denial and pushed “white genocide” conspiracy theories.

When we use AI, we are drinking from a well others have already poisoned.

The Qur’anic response is unambiguous, and found in a verse I regularly return us to; “O you who believe, be persistently standing firm in justice, witnesses for Allah, even if it be against yourselves” (Surah An-Nisa 4:135).

Standing firm in justice, in an age of algorithms, means refusing to treat AI output as objective truth, naming its biases openly, and being especially careful before letting AI describe, judge or represent people, especially women and minoritised communities.

My hallucinated poster was an irritation; the same technology deciding a woman’s CV or a Muslim’s visa application is an injustice, but they are both born of the same technology.

And the bias does not stay on our screens; it walks the streets with us, mounted on police vans. Facial recognition technology has been repeatedly shown to discriminate against people of colour (The Guardian, 2025). The foundational research is damning: Joy Buolamwini and Timnit Gebru’s landmark Gender Shades study found that commercial facial analysis systems misclassified darker-skinned women at rates approaching 35%, while making almost no errors on white men, and a US government (NIST) evaluation of nearly 200 algorithms confirmed false-positive rates for Black faces that were, in some systems, ten to a hundred times higher than for white faces.

In the US, this has already produced wrongful arrests of Black men misidentified by a machine. And this flawed technology is being scaled up, in Britain included, as live facial recognition spreads across public spaces with minimal legislation governing it. A surveillance state built on biased tools does not surveil everyone equally: it watches Black, brown and Muslim communities hardest, laundering old profiling through new mathematics.

For communities already familiar with being watched, from Prevent referrals to mosque surveillance, this is not an abstract civil-liberties debate. It is our faces in the database.

AI and Palestine

There is one place where every thread of this article; bias, surveillance, data, dehumanisation, converges with the most terrible clarity, and no Muslim discussion of AI ethics can look away from it: Palestine.

Since October 2023, investigations by journalists and human rights organisations have documented the Israeli military’s use of AI-driven systems in its assault on Gaza. Human Rights Watch (2024) examined four such digital tools. One, which the military calls “The Gospel,” algorithmically generates lists of buildings to bomb, including so-called “power targets,” civilian structures attacked, according to intelligence sources quoted in media reports, to shock the population (The Guardian first reported on The Gospel in December 2023). Another, “Lavender,” uses machine learning to assign the residents of Gaza numerical suspicion scores grading human beings by algorithm for their likelihood of being marked as targets. A third, chillingly named “Where’s Daddy?”, reportedly tracks phones to alert operators when a marked man arrives at a particular location–often his family home–so he can be attacked there, among his children. A fourth monitors evacuations through mass mobile phone tracking.

Human Rights Watch found these tools rely on faulty data, inexact approximations and sweeping surveillance of Gaza’s entire population, and warned that their use risks committing (more) war crimes. Israeli officers have said openly that these systems let them generate in days the number of targets that once took a year.

Nor is this the military’s project alone. In her policy brief for Al-Shabaka, the Palestinian digital rights analyst Marwa Fatafta (2025) traces how the infrastructure beneath these systems is supplied by the biggest names in the tech industry. Through the $1.2 billion Project Nimbus contract, Google and Amazon provide the Israeli government and military with cloud and machine-learning capability. Leaked documents reported by the Guardian revealed Microsoft’s cloud hosted a mass surveillance programme storing millions of intercepted Palestinian phone calls; Microsoft’s own review later confirmed a unit of the Israeli defence ministry had used its services for prohibited surveillance, and it suspended some services. Fatafta’s conclusion is blunt:

“AI for good” branding has collapsed in Gaza, where commercial AI now underpins what she names as complicity in grave violations of international law.

Human Rights Watch gives this process a name that should haunt every one of us who types casually into AI tools: digital dehumanisation; the reduction of human beings to data points, surveilled, sorted and scored, until it becomes easier to decide who may be harmed. Our tradition has always known where that road ends. “Whoever kills a soul…it is as if he had slain humankind entirely” (Surah Al-Ma’idah 5:32). Accountability before Allah cannot be outsourced to an algorithm.

What does this mean for us, ordinary users in London or Lagos or Lahore? At minimum: awareness, refusing to let this be buried while we argue about posters; bearing witness, because the Qur’an commands us to stand as witnesses for justice even when it is inconvenient (Surah An-Nisa 4:135); scrutinising where our money and data go, supporting the tech workers, rights groups and accountability campaigns Fatafta describes; and holding the full picture whenever we evaluate this technology.

The same machine learning that restored my CapCut project is a cousin of the machine learning that scores human beings for death. And that’s the uncomfortable truth of it.

It is precisely here that the journalist Karen Hao offers us a way to think this through and act rather than wallow in despair. In Empire of AI (2025) Hao argues that AI’s power players operate as a new kind of empire: laying claim to the work of artists and writers, the data of billions, the labour of moderation workers in Kenya and India paid a dollar or two an hour to sift traumatic content, and the land, water and energy of communities from Chile to the American South. Gaza is where that imperial logic reaches its most lethal point. But Hao’s reporting carries a second thesis, and it is the one I want us to hold onto:

empires feel inevitable, yet every empire in history has fallen, and they fall when broad coalitions of ordinary people refuse them.

She points to resistance that is already working: data centre protests that stalled over $100 billion worth of facilities in 2025 alone, tech workers refusing military contracts, communities defending their water. And she has turned that evidence into infrastructure, launching the AI Resist List; a global, community-built record of movements holding the industry accountable explicitly to counter Silicon Valley’s inevitability narrative with living proof that people are pushing back and winning. 

For us as Muslims, none of this should sound foreign. Ours is a tradition that commands whoever sees a wrong to change it with their hand; if unable, then with their tongue; if unable, then to reject it in their heart, “and that is the weakest of faith.” (Muslim)

Data is an Amānah

Then there is what we ourselves feed these AI tools. Every prompt may be stored, analysed and used to train future models. The Qur’an forbids spying and violating private matters (Surah Al-Hujurat 49:12), and the Prophet ﷺ taught that gatherings are held in trust. 

When we paste a friend’s messages into AI for advice on how to respond, or upload a community member’s personal situation, we may be handing someone else’s protected private matters to a corporation’s servers without their knowledge or consent. And the trust extends to creators: generative image tools are trained on the work of real artists, overwhelmingly without permission or payment.

Every slop poster is built on that unacknowledged labour which is one more reason it can never be the “harmless shortcut” it pretends to be. 

And then there are the endless “trends” that seem cute, fun and innocuous but are in fact great ploys not only at gathering more data on us but also making us more self-centred. We’ve seen the AI “Barbie-box challenge” where we all turned ourselves into boxed dolls, or the latest trend of saying the same phrase in various registers from sarcastic to flirtatious. Remember the “10-Year Challenge” which swept social media in 2019? Technology writer Kate O’Neill warned in Wired that millions of users were unwittingly assembling a perfectly labelled dataset (pairs of faces a verified decade apart) of exactly the kind needed to train facial recognition algorithms on age progression. 

The same logic applies to today’s trends: when OpenAI’s Ghibli-style image generator went viral in 2025, privacy experts cautioned that users were voluntarily handing over fresh biometric data, and, because the uploads were voluntary, it sidestepped the data protection hurdles that usually apply. And the latest round of voice trends carry another parallel risk: researchers have shown that as little as three seconds of audio is enough to clone a voice. There is no proof that any of these trends was deliberately engineered to harvest data, but that is beside the point. Whether by design or not, participating in them donates precisely the curated faces, voices and family photographs these systems need, under terms most of us never read.

When we upload the faces of our friends, family and children to be “cartoonified,” we are surrendering not our own data but theirs given to us in trust. 

Ethical Muslim AI use means treating data as amānah: never inputting others’ personal information without permission, being cautious with our own, and remembering that once uploaded, an image can never really be fully erased. 

The Sacrilege of Water

Finally, the consideration I see discussed least in our circles, every AI query has a physical cost, and it relates back to Abdurrahman’s third existential covenant: stewardship of creation. Data centres consume staggering quantities of drinking water (yes, DRINKING water), primarily for cooling. Researchers estimated that training GPT-3 consumed around 700,000 litres of freshwater for on-site cooling alone, and some 5.4 million litres once electricity generation is included, projecting global AI-related withdrawals of 4.2–6.6 billion cubic metres annually by 2027 (Li et al., 2023).

A UN University report this year warned data centre water consumption could reach 9.3 trillion litres by 2030; that’s enough to cover the basic annual water needs of over a billion people. Large data centres can use up to five million gallons of water a day, according to the Environmental and Energy Study Institute. And here in the UK, the Environment Agency has flagged a projected shortfall of nearly five billion litres of water per day by 2050, with AI-driven demand escalating the pressure. Our tradition speaks directly to this. “Eat and drink, but do not be excessive; indeed, He does not love the wasteful.” (Surah Al-A’raf 7:31) 

The Prophet ﷺ rebuked excess in water even for wudu, even “if you were on the bank of a flowing river” (Musnad Ahmad). If wastefulness is forbidden at a riverbank, what of ten idle image generations for a poster nobody asked for?

This, too, was in my viral post, and I stand by it: the environmental cost belongs squarely inside our fiqh of daily AI use.

Moving Forward

So where does this leave an AI-sceptic who was once rescued by the very tool she distrusts?

It leaves me here, with a working ethic I can hold on to. Verify everything, whether that is religious content or otherwise, before acting on it or passing it on (tabayyun). Never present AI-generated content, whether faces, voices, reminders or posters, as real without disclosure, and never touch another person’s likeness without their consent (sidq, and basic adab). Interrogate every output for gendered, racialised and Islamophobic bias (‘adl). Guard your own data and other people’s as a trust, and honour the artists whose unpaid labour built these tools (amānah). Block the means to harm before the harm arrives, even when it makes you unpopular with event organisers (sadd al-dharā’i), whether that’s by refusing to be platformed by such events or refusing to attend them. Bear witness against the weaponisation of these systems, from the police van to Gaza, and refuse quiet complicity in them (shahādah, in its widest sense). And use these tools sparingly and purposefully, remembering the water beneath every answer (khilāfah, and the prohibition of isrāf).

AI restored my video that day, and I was grateful; then AI put a stranger’s face on my event poster, and I walked away. I believe both responses were correct, because gratitude for a tool is not surrender to it, and refusal of misuse is not an overreaction. Between the technocrats’ false prophecy of inevitability and a blanket rejection lies the path we were always called to; the middle one.

It’s ok to be slower and have less output if it means we can better fulfil our responsibilities towards God, knowledge, and creation.

The empires of AI are betting that we are too tired, too dazzled, too complacent or too busy to choose intentionality, resistance and accountability. We are the vicegerents of Allah on this earth, responsible for the care of the planet and all who/that inhabit it and on the Day we are asked about our deeds, “AI was inevitable” will not be an answer.