AI and Islam: What Islamic Ethics Says About AI

In May 2025, the International Islamic Fiqh Academy — the collective jurisprudential body of the Organisation of Islamic Cooperation — met in Doha and issued the first major cross-madhhab ruling on artificial intelligence.

Its verdict was not what many expected. The default ruling on developing and using AI, the Academy held, is permissibility — ibāḥah — subject to six binding conditions.

That single resolution reframes the whole conversation. The question is no longer “is AI halal or haram?” It is: what conditions must be met, and who is accountable when they are not?

This article sets out what Islamic ethics actually says about AI — the formal rulings issued so far, the questions of moral responsibility and consciousness, the specific matter of AI-generated fatwas, and a working framework drawn from the maqāṣid al-sharīʿah. Every source is linked.

Quick Answer

The default ruling is permissibility (ibāḥah). The International Islamic Fiqh Academy’s Resolution No. 258 (3/26), issued in Doha in May 2025, holds that developing and using AI is permissible subject to six ethical and legal conditions.
The ruling follows the purpose. Under the classical maxim “the means take the same ruling as the goals,” AI used for something lawful is lawful; AI used for something prohibited is prohibited.
AI is not morally accountable. Taklīf — legal responsibility — requires intellect, free will and accountability before Allah. A machine has none. The human being who builds, deploys or uses it carries the responsibility.
AI cannot be relied on for fatwas. Both Egypt’s Dar al-Ifta and IslamQA have ruled that it is impermissible to take religious rulings from AI applications, though a qualified researcher may use it to locate sources.
Fifty-three Muslim-majority states have adopted an AI ethics charter. ICESCO’s Riyadh Charter on AI in the Islamic World sets out shared principles including privacy, human dignity, fairness and accountability.
Consciousness is a separate question from capability. The Qur’an places the rūḥ beyond human knowledge (al-Isra 17:85), which is where Islamic thought locates the limit of the claim that a machine could ever be a person.

The 2025 Ruling: What the Fiqh Academy Actually Decided

The International Islamic Fiqh Academy is not a single scholar or a single school. It is the OIC’s collective body, drawing jurists from across the Muslim world, and its resolutions carry the weight of ijtihād jamāʿī — collective legal reasoning.

At its twenty-sixth session in Doha (4–8 May 2025), it resolved that AI is “a modern technology based on programs and machines that simulate human intelligence,” bringing “numerous benefits, though it is not without potential harms.”

Its ruling: the default is permissibility, subject to six conditions.

#	Condition
1	The purpose of its creation, use, funding and outcomes must be lawful
2	It must serve to bring benefit and prevent harm
3	It must not insult or abuse beliefs, religions or religious symbols
4	It must protect information and preserve public and private rights and freedoms
5	It must not threaten individual, societal or national security
6	It must adhere to honesty, proper documentation and transparency in its use

Read those conditions carefully and something becomes clear: they are almost all about the human beings around the system, not the system itself. Purpose, funding, honesty, transparency, protection of rights — these are duties on developers, financiers and users.

The Academy also referred one question forward for further study: whether AI can be granted legal personality. It did not answer it. That is an honest acknowledgement that the hardest question is still open.

The Starting Principle: Permissibility, and the Means

The Academy’s conclusion was not novel reasoning. It applies a foundational principle of Islamic law: al-aṣl fī al-ashyāʾ al-ibāḥah — the default ruling on things is permissibility, unless a text prohibits them.

Egypt’s Dar al-Ifta grounds this in the Qur’an directly:

وَسَخَّرَ لَكُم مَّا فِي السَّمَاوَاتِ وَمَا فِي الْأَرْضِ جَمِيعًا مِّنْهُ ۚ إِنَّ فِي ذَٰلِكَ لَآيَاتٍ لِّقَوْمٍ يَتَفَكَّرُونَ

“And He has subjected to you whatever is in the heavens and whatever is on the earth — all from Him. Surely in this are signs for people who reflect.” — al-Jathiyah 45:13

Note how the verse ends: for people who reflect. The subjection of the natural world to human use is tied, in the same breath, to the obligation to think about it.

From there, Dar al-Ifta applies the maxim “the means take the same ruling as the goals” (lil-wasāʾil ḥukm al-maqāṣid), citing al-ʿIzz ibn ʿAbd al-Salam. AI is a wasīlah — an instrument. A scalpel is not halal or haram; what you do with it is.

This is why asking “is AI haram?” is the wrong question. It is like asking whether writing is haram. The ruling attaches to the use.

Can a Machine Be Morally Responsible?

This is where Islamic ethics gives a sharper answer than most secular frameworks manage.

In Islamic law, moral and legal responsibility is called taklīf, and it does not attach to everything. It requires a specific set of conditions: being of sound intellect (ʿaql), having reached maturity, and possessing genuine choice (ikhtiyār). This is why children, the insane and the coerced are not held accountable in the same way.

The Qur’an ties responsibility directly to capacity:

لَا يُكَلِّفُ اللَّهُ نَفْسًا إِلَّا وُسْعَهَا ۚ لَهَا مَا كَسَبَتْ وَعَلَيْهَا مَا اكْتَسَبَتْ

“Allah does not require of any soul more than what it can afford. All good will be for its own benefit, and all evil will be to its own loss.” — al-Baqarah 2:286

The verse speaks of a nafs that earns — that acquires moral weight through its own acts. A language model does not earn anything. It has no intention (niyyah), no fear of accountability, no capacity to repent.

The practical consequence is significant and it cuts against a convenient modern excuse. “The algorithm decided” is not a defence in Islamic ethics. If an AI system denies someone a loan unjustly, spreads a slander, or kills, the moral weight falls on the humans in the chain — the designers who chose the objective, the company that deployed it, the operator who used it. Responsibility cannot be laundered through a machine.

This is also why the Fiqh Academy left legal personality for further study. Granting AI legal personhood is a question about liability allocation, not about whether the machine has a soul.

Consciousness, the Rūḥ, and the Limits of the Claim

Could an AI ever be conscious? Islamic theology approaches this from an unusual angle: it declines to speculate.

وَيَسْأَلُونَكَ عَنِ الرُّوحِ ۖ قُلِ الرُّوحُ مِنْ أَمْرِ رَبِّي وَمَا أُوتِيتُم مِّنَ الْعِلْمِ إِلَّا قَلِيلًا

“They ask you ˹O Prophet˺ about the spirit. Say, ‘Its nature is known only to my Lord, and you ˹O humanity˺ have been given but little knowledge.'” — al-Isra 17:85

The rūḥ is explicitly placed outside the domain of human knowledge. Whatever consciousness is, the Qur’an does not present it as an engineering problem.

What Islam does affirm plainly is the dignity of the human being:

وَلَقَدْ كَرَّمْنَا بَنِي آدَمَ وَحَمَلْنَاهُمْ فِي الْبَرِّ وَالْبَحْرِ وَرَزَقْنَاهُم مِّنَ الطَّيِّبَاتِ وَفَضَّلْنَاهُمْ عَلَىٰ كَثِيرٍ مِّمَّنْ خَلَقْنَا تَفْضِيلًا

“Indeed, We have dignified the children of Adam, carried them on land and sea, granted them good and lawful provisions, and privileged them far above many of Our creatures.” — al-Isra 17:70

That dignity — karāmah — is conferred, not earned through capability. Which means it does not transfer to a system merely because the system becomes more capable. A model that writes better poetry than you does not thereby acquire your standing before Allah.

The honest position is this: Islamic sources do not rule out machine consciousness as a metaphysical impossibility, because they do not engage the question. What they establish is that human dignity is not a function of intelligence, so the argument “it is as smart as us, therefore it is one of us” does not follow on Islamic premises.

A Framework: AI and the Maqāṣid al-Sharīʿah

The most useful tool Islamic ethics offers for evaluating a specific AI system is the maqāṣid — the five necessities that Islamic law exists to protect, as systematised by al-Ghazali and al-Shatibi.

Necessity	AI as protection	AI as threat
Dīn (faith)	Quran learning tools, translation, accessibility for the disabled	Unverified AI fatwas, fabricated hadith, simulated scholarly authority
Nafs (life)	Medical diagnostics, drug discovery, road safety, disaster response	Autonomous weapons, unsafe medical automation
ʿAql (intellect)	Tutoring, research acceleration, access to knowledge	Cognitive outsourcing, deskilling, engineered addiction to feeds
Nasl (lineage/family)	Child safety tooling, health screening	Deepfake abuse imagery, synthetic intimacy replacing family bonds
Māl (wealth)	Fraud detection, efficiency, financial inclusion	Algorithmic exploitation, opaque credit denial, mass job displacement

The value of this framework is that it forces a per-system judgement rather than a verdict on “AI” as an abstraction. The same underlying technology sits on both sides of every row.

The Fatwa Question: Can You Ask AI About Your Religion?

This is the most consequential practical question, and the scholarly answer has been notably consistent.

Egypt’s Dar al-Ifta, in a fatwa issued 2 December 2025 by Grand Mufti Dr. Nazir Mohammad Ayyad, ruled: “it is impermissible in Islamic law to rely on AI applications to obtain a fatwa.”

The reasoning is not technophobia. It is about what a fatwa is. A muftī must know the sources, the abrogating and abrogated texts, the language, and — critically — the questioner’s actual circumstances. Dar al-Ifta cites al-Qarafi: a ruling built on a custom changes when the custom changes. AI, it holds, cannot assess “the shifting contexts of time, place, and the particular circumstances surrounding each fatwa.”

IslamQA reached the same conclusion in July 2025, and added a technical diagnosis that any AI practitioner will recognise: the sources are unknown, the training data may be “deviant or distorted,” the output can be changed by rephrasing the question in ways that should not affect the ruling, and the systems produce “names, numbers and statistics which appear to be accurate and objective, but in reality are fabricated” — what it calls, with the exclamation mark in the original, “AI hallucinations!”

Both distinguish carefully between two users:

A qualified scholar or researcher may use AI to locate sources, summarise material and map an issue — then must return to the original sources and verify everything himself. This is legitimate, because the means take the ruling of the goal.
An ordinary Muslim may use it to get a general orientation, but must not act on a ruling it generates. IslamQA’s practical advice is sound: use it as “an intelligent search engine” — ask it for references and links, then go read the actual fatwa at its source.

There is a genuine minority position, which IslamQA records fairly: some contemporary researchers distinguish universal rulings (what breaks wudu’, what invalidates a fast) from circumstance-dependent ones, and permit AI for the former where the app draws only on a trustworthy institution’s own verified fatwas. IslamQA considers the stricter view stronger. Both views agree that a general-purpose chatbot is not a source of religious rulings.

Justice, Bias and the Algorithm

Algorithmic bias is discussed in secular AI ethics as a fairness problem. In Islamic terms it is a matter of ʿadl — justice — which the Qur’an commands unconditionally and specifically warns must not bend to group animus (al-Ma’idah 5:8, an-Nahl 16:90).

This maps onto the Fiqh Academy’s second and fourth conditions. A hiring model that systematically disadvantages a group, or a credit system whose refusals cannot be explained, fails the requirement to “prevent harm” and to “preserve both public and private rights.”

The Academy’s sixth condition — honesty, proper documentation and transparency — is striking in how closely it tracks what AI governance calls explainability. A system whose decisions cannot be accounted for is difficult to reconcile with a tradition in which the person who acts must be able to answer for the act.

Privacy, Surveillance and Amānah

Islamic law arrived at a strong privacy principle centuries before data protection existed, and it is stated as a direct prohibition:

يَا أَيُّهَا الَّذِينَ آمَنُوا اجْتَنِبُوا كَثِيرًا مِّنَ الظَّنِّ إِنَّ بَعْضَ الظَّنِّ إِثْمٌ ۖ وَلَا تَجَسَّسُوا وَلَا يَغْتَب بَّعْضُكُم بَعْضًا

“O believers! Avoid many suspicions, ˹for˺ indeed, some suspicions are sinful. And do not spy, nor backbite one another.” — al-Hujurat 49:12

Lā tajassasū — do not spy. Not “do not spy without cause”; the prohibition is stated flatly, alongside suspicion and backbiting, as a violation of another person’s standing.

Modern AI is, in large part, a technology of inference from personal data. Facial recognition, behavioural prediction, emotion detection and profiling all sit uncomfortably close to tajassus. And data held about another person is an amānah — a trust:

إِنَّا عَرَضْنَا الْأَمَانَةَ عَلَى السَّمَاوَاتِ وَالْأَرْضِ وَالْجِبَالِ فَأَبَيْنَ أَن يَحْمِلْنَهَا وَأَشْفَقْنَ مِنْهَا وَحَمَلَهَا الْإِنسَانُ

“Indeed, We offered the trust to the heavens and the earth and the mountains, but they ˹all˺ declined to bear it, being fearful of it. But humanity assumed it.” — al-Ahzab 33:72

The verse is about the weight of moral responsibility itself. Applied here: a company holding a billion people’s behavioural data is holding an amānah, whether it uses that word or not.

Deepfakes and the Duty to Verify

Generative AI makes fabricated evidence cheap. Islamic law has a directly applicable rule:

يَا أَيُّهَا الَّذِينَ آمَنُوا إِن جَاءَكُمْ فَاسِقٌ بِنَبَإٍ فَتَبَيَّنُوا أَن تُصِيبُوا قَوْمًا بِجَهَالَةٍ فَتُصْبِحُوا عَلَىٰ مَا فَعَلْتُمْ نَادِمِينَ

“O believers! If an evildoer brings you any news, verify ˹it˺ so you do not harm people unknowingly, becoming regretful for what you have done.” — al-Hujurat 49:6

Fatabayyanū — verify. The Muslim tradition of isnād, the chain of transmission that underpins the entire hadith sciences, is arguably the most rigorous information-verification system any civilisation has produced. It exists precisely because a claim’s source determines its weight.

An AI-generated video has no isnād. Neither does an AI-generated hadith — and models do fabricate them, complete with plausible-looking book and number references. For anyone producing Islamic content, the rule follows directly: verify every narration at its source before publishing it. A confident-sounding citation is not a citation.

Work, Automation and the Dignity of Labour

Displacement is the AI question most people feel personally. Islam attaches real honour to earning by one’s own effort:

The Prophet ﷺ said: “Nobody has ever eaten a better meal than that which one has earned by working with one’s own hands. The Prophet of Allah, Dawud, used to eat from the earnings of his manual labour.” — Sahih al-Bukhari 2072

That the example given is Prophet Dawud — a king who made armour with his own hands — is not incidental. Work is dignified regardless of whether one needs the income.

This does not make automation prohibited; labour-saving tools have always been used. But it does mean that under the maqāṣid, displacing human work at scale is not a neutral efficiency question. It touches māl and human dignity together, and the Academy’s requirement to “bring benefit and prevent harm” applies to the second-order effects, not only the product.

Using AI Well: A Practical Checklist

Drawing the rulings together, for an ordinary Muslim using these tools:

Never take a religious ruling from a chatbot. Ask it for sources, then read the source. For anything that depends on your circumstances, ask a qualified scholar.
Verify before you transmit. Do not share an AI-produced hadith, verse reference or statistic without checking it at Quran.com, Sunnah.com or a recognised institution.
Check the purpose. The ruling follows the use. Ask what the output is for before asking whether the tool is permissible.
Guard other people’s data. Do not feed private information about others into systems you do not control.
Do not counterfeit a person. Voice cloning, deepfakes and impersonation engage prohibitions on deception and slander regardless of the technology.
Watch what it does to your own ʿaql. A tool that thinks for you erodes the faculty Islam treats as the basis of accountability.

Learning at the Source

There is a reason the scholars keep returning to the same point: religious knowledge is transmitted, not generated.

The angels’ response when taught by Allah is the model of intellectual humility in the Qur’an:

قَالُوا سُبْحَانَكَ لَا عِلْمَ لَنَا إِلَّا مَا عَلَّمْتَنَا ۖ إِنَّكَ أَنتَ الْعَلِيمُ الْحَكِيمُ

“They replied, ‘Glory be to You! We have no knowledge except what You have taught us. You are truly the All-Knowing, All-Wise.'” — al-Baqarah 2:32

AI can help you find a book. It cannot correct your makhraj, hear the difference between a madd held for two counts and four, or notice that you have been rushing your recitation for a month. That requires a teacher.

Al-Ilm Academy offers one-to-one Quran and Tajweed classes with qualified teachers, and the first trial class is free — no obligation. Use the tools for what they are good at, and learn your Qur’an from a human being with an ijāzah.

Frequently Asked Questions

Is artificial intelligence haram in Islam? No. The International Islamic Fiqh Academy ruled in Resolution No. 258 (3/26), issued May 2025, that the default ruling on developing and using AI is permissibility (ibāḥah), subject to six conditions covering lawful purpose, benefit and harm, respect for religion, protection of rights and information, security, and transparency. AI used for a prohibited purpose takes the ruling of that purpose.

Can I ask ChatGPT or another AI for a fatwa? No. Egypt’s Dar al-Ifta ruled in December 2025 that it is impermissible to rely on AI applications for fatwas, and IslamQA reached the same conclusion in July 2025. AI does not meet the conditions of a muftī, cannot weigh your circumstances, and fabricates sources. You may use it to find references, then read them at their origin.

Can AI have a soul or be conscious? The Qur’an places the nature of the rūḥ beyond human knowledge (17:85). Islamic sources do not engage machine consciousness directly. What they do establish is that human dignity (karāmah) is conferred by Allah (17:70) rather than earned through capability — so increased intelligence does not confer human standing.

Who is responsible if an AI system causes harm? The humans involved. Taklīf — legal accountability — requires intellect, maturity and free choice, none of which a machine possesses. Responsibility falls on those who designed, funded, deployed or used the system. The Fiqh Academy referred the separate question of AI legal personality for further study without deciding it.

Is it permissible to use AI to learn Quran or Islamic studies? Using AI as a study aid, translation tool or search assistant is generally permissible under the same default of ibāḥah, provided everything is verified against authentic sources. It should not replace a qualified teacher, particularly for recitation and Tajweed, which require live correction and a transmitted chain.

What is the Riyadh Charter on AI? An AI ethics framework issued through ICESCO and unanimously approved by its 53 member states, built on principles including integrity and fairness, privacy and security, reliability and safety, transparency and explainability, accountability, humanity, and social and environmental benefit. The Fiqh Academy reviewed it before issuing Resolution 258.

Is it haram to use AI to generate Quran recitation or Islamic images? This is a live question with no settled consensus, and it is more contested than general AI use. It engages the prohibition on deception, the requirement of isnād in transmitting religious content, and existing rulings on imagery. Anyone producing such content should consult a qualified scholar rather than rely on this article.