Fatwā and AI: A Literature-Based Assessment On How AI Impacts the Role of the Muftī

Author: Shoayb Ahmed
University of South Africa
Email: eahmeds@unisa.ac.za
DOI: 10.25159/2957-9163/14400

Abstract

This article explores the intersection of fatwā (Islamic legal opinions) and artificial intelligence (AI), particularly chatbot technology, to assess its impact on the role of the muftī, who traditionally issues fatwas. The rise of AI-powered chatbots presents both opportunities and challenges in fatwā issuance. AI offers advantages like accessibility, consistency, and the ability to handle a high volume of queries, potentially improving the speed and reach of religious guidance, particularly in remote areas. However, the lack of empathy and nuanced understanding in chatbots raises concerns, as they may fail to address the complexities of Islamic jurisprudence or provide contextually appropriate responses. A literature review of existing academic work on AI and fatwā was conducted. Some of these studies contain case studies of AI applications in religious guidance, content analysis of fatwā texts and AI algorithms, and comparative analysis of different AI approaches. The article assessed the challenges and opportunities for the muftī in an AI-dominated world. It examined the impact of AI on the muftī’s role and provides recommendations for adapting to this technology. It recommends the need to incorporate AI owing to, among other factors, the benefits of its large datasets. However, such fatwā provision must retain the human element, namely the traditional, physical role of the muftī, who is always alert and sensitive to context and can gauge what type of ruling is appropriate in each situation. A partnership between AI and human judgment has the potential to enhance the efficiency and accessibility of fatwā issuance while preserving its integrity.

Introduction

The rapid growth of digital and cyber religion, and the consequent rise of online religious platforms, religious chatbots, religious robots, and various artificial intelligence (AI) applications on religion, have fundamentally transformed the religious landscape (Yucel and Ismail 2021, 2). This also applies to the muftī and the fatwā-giving process. AI will almost certainly influence and alter how muftis are trained and how they will disseminate fatāwā in the future. Questions outnumber muftis (fatwā providers) who can give answers. Thus, this article argues that in light of the rapid development of AI and its impact on the religious landscape, AI-powered fatwā systems are an inevitable development in modern religious discourse.

This article investigated how the muftī will have to position themselves given the wide-ranging presence of AI. What follows is my reflection on, and development of, the points presented in the literature and my view of the future role of the muftī given these developments. The methodology is primarily a literature review with some articles concentrating on case studies, others surveying existing literature on the topic, and others analysing the content. It approaches this problem by:

• Providing an overview of the literature relating to AI and fatāwā.
• Assessing, based on this literature, the challenges, and opportunities for the muftī in an AI-dominated world.
• Proposing, in conclusion, a recommendation of its own, namely, that the provision of fatwā needs to incorporate AI owing to, among other factors, the benefits of its large datasets. However, such fatwā provision must retain the human element, namely the traditional, physical role of the muftī, who is always alert and sensitive to context and can gauge what type of ruling is appropriate in each situation.

AI and Fatāwā: An Overview of the Literature

The literature on AI and fatwā can be divided into three categories:

• Literature describing the state of the field concerning AI and fatwā
• Literature discussing the problems and prospects of producing fatāwā using AI
• Technical literature that produces models for generating fatāwā using AI

Each of these is dealt with in turn. It should be noted that what follows is an analysis of a representative sample of literature in these categories, not an exhaustive survey of all the literature that has been produced in relation to the topic.

Literature Describing the State of the Field Concerning AI and Fatwā

Singer (2021) and Tsourlaki (2022) have made important contributions describing the state of AI concerning fatwā provision, particularly in the Arab world.

Singer primarily looked at how muftis have responded online to questions regarding robotics and artificial intelligence. For this purpose, she selected the Qatar-based website, Islamicweb.net, and analysed 14 fatāwā from that location between 2002 and 2019, drawing a comparison between how questioners and muftis dealt with the two domains. There were more questions on robotics and these, in sum, dealt with the following: whether making robots was permitted in Islam; whether they should resemble human beings or not; the types of activities for which they can be employed; and the philosophical question of whether the human being is creating something that should only be done by God himself. The response of the muftis included the following: that robots could be manufactured since, under the well-known principle of Islamic law, all things that are not expressly forbidden by that law are permitted; that they cannot resemble human beings; that they may be used for all types of permitted activities; and that a human being cannot innately create something new and so they are not taking God’s place.

There were fewer questions concerning AI—four as opposed to ten on robotics, perhaps because LLMs (Large Language Models) had not emerged as a phenomenon by 2019—but some of these interacted with robotics, namely whether it was permitted, and whether AI can challenge God’s uniqueness. Again, the responses were the same, namely that in principle AI is permitted and that it cannot intrinsically challenge God’s uniqueness. However, rather curiously, questioners also asked whether strong AI is possible, that is, whether machines would be able to think like humans. The muftis did not delve into the technical distinction between strong and weak AI; however, they did not hold that a machine can go beyond its inbuilt capabilities in the first place. In general, though, they appeared reluctant to issue a detailed opinion about artificial intelligence. This may be because they wanted to see how AI technology develops and its subsequent impact, or because they were not familiar with the technical aspects and hence could not commit to issuing a verdict on something that still seemed obscure and inconclusive.

In her conclusion, Singer makes an important observation that there is a disjunct between the concerns of the questioner and the replies of the muftis on the question of AI. The questioners show a concern with strong AI (and so the hype surrounding the issue), while the muftis are conservative and less speculative (and so probably more technically accurate) in their responses to this concern. While the muftis may have been technically accurate—AI in its current state is still weak AI—their responses were sparse in detail and based on general principles of Islamic law, as duly noted by Singer. This indicates that they need to research this issue in a far more rigorous way if they are to speak to the concerns of the masses. This is essential if muftis are to navigate the widening scope of the AI landscape.

Tsourlaki’s 2022 study dealt with AI-based fatwā provision models at two institutions, the Dubai Ministry of Religious Affairs and the Dar al Iftā in Egypt. Her research was conducted between August 2020 and February 2022.

In the case of the Dubai Ministry of Religious Affairs in the United Arab Emirates (UAE), the AI system employed responded to a question by identifying a previously issued fatwā already recorded and stored in its database. The choice of the fatwā was based on the semantic proximity between the question and the chosen fatwā, which the AI system calculated based on its linguistic relevance. However, this purely automated dispensing of fatāwā was unsuccessful. By March 2022, the UAE model was deactivated, and the service was no longer available. Tsourlaki hazards the guess that this may have happened because of the low number of users who opted for such a service (2022, 13).

On 1 January 2020, a few months after Dubai’s announcement, the Al-Azhar Global Fatwā Centre in Cairo announced the design of its own AI fatwā system. Its main purpose seemed to be to “monitor abnormal and extremist fatwās.” The Egyptian plan was expected to operate in three stages. The first stage consisted of collecting and examining previously issued fatāwā and storing those deemed legitimate and acceptable. In the second stage, these approved fatāwā are available to the questioners according to the lexical proximity of the users’ questions. The third step of the Egyptian model is a “follow-up process.” In this process, a muftī will be tasked to contact the mustaftī by employing each user’s collected personal details and ensuring that the fatwā provided fulfils the need and answers the question.

While both governments would be careful about monitoring fatāwā to ensure they do not harm the interest of the state, there is an important distinction between the two AI processes: in the Dubai case, it is fully automated; in the Egyptian case, the physical role of the muftī as the final overseer of the automated fatwā is maintained. This, of course, allows the possibility that they may not approve the fatwā in each context. I believe that this is an important overseeing role: a fatwā, by its nature, responds to a specific context and an assessment of that context can only be done by a muftī. Despite the widening scope of AI, the physical role of the muftī remains critical.

It appears that Tsourlaki’s interviewees were more favourable to the Egyptian approach. She conducted surveys in 2020 during which she questioned people about AI-related fatāwā. Some of her findings were that 84.5% of the participants used a fatwā online at some point in their lifetime by accessing information on Islamic websites. More precisely, 45.5% used fatāwā daily, weekly, or monthly. In most cases, these people were using online repositories. When the participants were asked about whether it was important to know if the fatwā was by a human or a computer, 92.7% felt that it was important. However, 96.3% would not trust a fatwā generated by a computer, with 61.8% opting to reject such a fatwā completely. This may partly be the reason why by March 2022, the UAE “Virtual Ifta” system seemed to come to an abrupt end. It seems to have received very little coverage and very little academic interest from Muslim scholars.

But, as mentioned, Tsourlaki’s study also alerts us to the role of the state. By employing AI to monitor and control the fatāwā disseminated worldwide, religious institutions and the political powers behind them might be able to control religious discourse in a very substantial way, perhaps in a manner unprecedented in the history of the Islamic faith. The beliefs and reasoning of only one school of thought (madhhab) within Islamic jurisprudence (fiqh) and the societal circumstances of only one country may dictate the lifestyle and obligations of every Muslim worldwide. The standardisation of fatāwā through AI, which is a logical development after the standardisation of sermons delivered in mosques—already implemented in many Muslim-majority countries—will signify the further attenuation of the independence of the muftī. The muftī (one who issues an informed legal opinion) might unwittingly become the qāḍī (judge)—one who can enforce a particular legal ruling. Historically, muftis operated independently, while the qāḍī operated under the jurisdiction of the court and was appointed by the state. The muftī issues a fatwā and rules in devotional acts of worship, which is not the realm of the qāḍī. The qāḍī rules in court, while the muftī may rule at his home. The jurisdiction of the qāḍī is different to that of the muftī. See al-Furuq by al-Qarafi (n.d.), 4/48–52; Qawaid al-Ahkam fi Masalih al-Anaam by Abdussalam (n.d.), 1/120, and al-Tamyiz by al-Qarafi (1995, 46–50).

There have been initiatives regarding fatwās based on technology in other parts of the Muslim world. For example, JAKIM, which is the main authority in charge of Islamic religious affairs in Malaysia, started an initiative to introduce an e-Fatwā website in 2009. The website serves to gather all the fatwās issued by fatwā institutions in Malaysia. e-Fatwā websites and mobile applications were operational in 14 states in Malaysia (Mohd et al. 2018). The Dubai Ministry of Religious Affairs, which is part of the Islamic Affairs and Charitable Activities Department (IACAD) in the government, has an office responsible for issuing fatwās. https://dubai.ae/religious-affairs Dar al-Ifta, Egypt, was founded in 1895 and is affiliated with the Ministry of Justice. Its function is to issue fatwās for the people of Egypt, and it also serves the global Muslim community and incorporates various global fatwā bodies. https://www.dar-alifta.org/en/about

In addition to Singer and Tsourlaki’s studies, we may also point to one by Wahid and Kususiyanah (2020), which gives insight into the state of the field. Speaking from an Indonesian context, the authors discuss several points regarding the application of AI and machine learning in legal analysis, demonstrating how technology has transformed the practice of fatwā, moving from traditional book-based research to electronic databases and AI-driven tools that enhance the efficiency of legal research. The study found that such resources were able to effectively and efficiently analyse legal documents, thus aiding the understanding of legal relationships, and supporting the transformation of legal texts into actionable regulations.

Literature Discussing the Problems and Prospects of Producing Fatāwā Using AI

Articles by Sa’ad et al. (2020), Yahya and Junaid (2021), Polat et al. (2022), Ghaly (2023), Latifi (2022), and Kausar et al. (2024) all concur that AI could speed up the fatwā provision process and that it would certainly serve as a database and bank of information from which the muftī can draw. It would make the fatwā provision process more efficient and cost-effective and make the information more accessible. Some authors particularly refer to its effective usage in Islamic finance. But they all identify certain challenges. Each of their articles will be discussed individually.

Sa’ad et al. (2020) investigated the potential Sharī‘ah and regulatory issues surrounding robo-advisors. Robo-advisors have been defined as online wealth management services that offer computerised, algorithm-based portfolio management without using human financial planners. They have already had a considerable impact on the financial industry. In terms of Islamic finance, the authors appear to suggest that robo-advisors can also act as “smart muftis” through input of the relevant information derived from Islamic law. In this way, they can serve as assistants to physical Sharī‘ah advisors. For the authors, “smart muftis” might be able to evaluate the Sharī‘ah sources and the existing rulings across the various schools of Islamic jurisprudence and may suggest the best practices chosen by Islamic scholars on a certain issue. They will also act as a database for collected fatāwā and facilitate training programmes in Islamic finance. However, smart muftis and robo-advisors will support, not supplement, muftis. Fully automated systems can produce misleading information; therefore, human supervision is needed. Thus their article, while pointing out the opportunities of AI in Islamic finance, still supports the traditional, indispensable role of the physical muftī.

Overall, the article presents a comprehensive view of how robo-advisors and smart muftis can enhance the Islamic finance industry by improving speed, compliance, efficiency, and service delivery while addressing the concerns of the Sharī‘ah. Their points are echoed in an article by Rhamadhi (2024), which is also focused on Islamic finance, and who adds that AI technology must be trained to link the fiqh (legal) rulings with the relevant ratio legis; that it must include the pronouncements from all major fiqh academies; and that the AI system must be updated regularly. However, one can clearly see that such automated systems as discussed by Sa’ad and Rhamadhi can be extended to other areas of Islamic law as well.

Yahya and Junaidi (2021) make the point that a scholar’s reputation and his or her mastery of the sources of Islamic jurisprudence remain crucial in the digital era. They also highlight what will be the increasingly important role of collective ijtihād (legal reasoning) in the digital era. Having a dataset that gathers information from a wide variety of sources and is accessible to scholars globally naturally facilitates such a process. Thus muftis need to be conversant with the potential of technology to facilitate more current ways of legal reasoning while still firmly grounded in the traditional strengths of their craft.

Ghaly (2023), discussing the increasing use of robots and AI-based technologies and its impact on ethics, makes the point that future Islamic ethical discussions in these emerging fields should benefit from the accumulated experience of other disciplines such as Islamic bioethics and not simply be confined to the legal sphere alone. He thus also feels that collective reasoning (al-ijtihād al-jamāʿī) will become more vital in the future. Ghaly more broadly argues that in order to develop Islamic ethical perspectives based on a proper understanding of the technical aspects of AI, Muslim religious scholars should also collaborate with engineers, computer scientists, and specialists in related fields. Such collaboration may be essential for the muftī in the provision of a fatwā.

In an important article on AI and ijtihād, Latifi (2022) investigates the potential challenges associated with employing generative artificial intelligence (AI) models in the process of Islamic, especially Shī‘ī, legal reasoning, discussing both its challenges as well as the opportunities it presents. Regarding opportunities: AI allows the muftī convenient access to a large database of fatāwā on which to base his or her decision; it allows for more accurate and detailed scrutiny of the ḥadīth sources on which a fatwā may be based; and it allows for more timeous, and even instantaneous responses to questions. Regarding the challenges of an AI fatwā system, Latifi presents a number of these: it may not know the context of a particular question or how a muftī may have changed their mind on a particular issue; it may lead to a lack of creativity and rigour from muftis who excessively rely on it; it may prove expensive technology to people in developing countries; its Natural Language Processing (NLP) may still be inadequate for lesser known languages used by questioners; an overreliance on AI can lead to social disconnection, in contrast to the physical connection to the muftī cultivated by the traditional fatwā provision process; an online AI fatwā may be subject to a country’s specific laws; AI fatāwā are also less transparent and subject to less accountability than a traditional physical fatwā; and, of course, it is subject to the well-known issue of AI bias and hallucination. Latifi also makes the crucial point that in generative AI the same question can yield more than one answer, not only differing in wording but, more importantly, in content. While it is common for muftis to change or alter their fatwā because of changing circumstances or being privy to additional information, this is not the case with generative AI. This is because it would generate new and different responses within seconds of the first, even though the circumstances may not have changed. Sometimes these different responses may even be inconsistent. Latifi’s article is a crucial direct intervention on how AI impacts the role of the muftī and we believe that it holds a lot of salience for fatwā provision processes in a Sunni context as well.

Technical Literature That Deals With Producing Models for Generating Fatāwā Using AI

The potential of AI to act as a fatwā provider has led Muslim computer scientists and engineers to develop various systems to bring this to fruition. Already in 2015, Elhalwany et al. attempted to develop a fatwā QA (question answering) system that uses a programming approach for text indexing, retrieval, and system learning. Additionally, the system sought to address language and domain challenges. Their SOPHIATCBR (SOPHIA [SOPHisticated Information Analysis] textual case-based reasoning framework) involved: 1. Case Knowledge Discovery; 2. Narrow Theme Discovery; 3. Similarity Knowledge Discovery; 4. Case Assignment Discovery; and 5. Internal Cluster Structure Discovery. What this basically meant was that a system would align a user’s input data to similarly themed cases on the dataset and generate the most appropriate response. The authors developed a dataset of approximately 5000 cases and worked in collaboration with the Darul Ifta, Egypt’s national fatwā issuing body.

Later, Munshi et al. (2021, 2022), now writing in the emerging age of Large Language Models, noted that with the emergence of social media and related technologies, there was a shift in how Islamic knowledge is disseminated. Their work focused on a chatbot system which could relieve the load of the muftis. Other advantages of such a chatbot would include categorising and directing questions to specialised experts, as well as providing immediate answers by matching users and their inquiries to a database of previous responses from users with similar questions, thus building on Elhalwany et al.’s (2015) work. They crucially noted that many of the references in Islam are still in Arabic and the Arabic language is considered among the Low-NLP Resources languages, unlike English. This called for the need for both TL (Transfer Learning) and MTL (Multi-Task Learning) to help solve this issue. There is a wide gap in applying the above techniques to Arabic NLP tasks. But they also noted that more recent, pre-trained models for Arabic like BERT and AraGPT have been released that assist in bridging this gap. Ultimately, the authors present the first steps towards building an automated fatwā system using AI and deep learning NLP methods, based on a large dataset of fatāwā, ultimately incorporating 85,000 questions and answers. For this dataset, they performed unsupervised topic modelling and analysis, with baseline models for both topic classification and QA. They evaluated their baselines in various aspects like the effect of sequence modelling and that of pre-trained embeddings. Crucially, their system also focuses on questions and answers in social media channels and sought applications for their system not only in Arabic but in the various languages in which questions are posed to muftis. Their system also took into account the background (origin, language, ethnicity, and sex) of the questioner in order to facilitate the process of an appropriate automated answer.

Both Elhalwany (2015) and Munshi et al. (2021, 2022) are acutely conscious of the limitations of fully automated fatāwā and assert that the traditional role of the muftī, especially given his or her knowledge of the context of a question, remains critical. The muftī remains the final overseer. So, both articles, written by authors whose specialisations are in scientific fields, show a deep awareness of traditional Islamic jurisprudential principles. However, the question should be asked: do traditional muftis show an equal awareness of the technical aspects underlying an automated AI fatwā system? It is clear from what has outlined thus far regarding the opportunities and challenges presented by such a system that they need to develop such an awareness.

Assessment: Challenges and Opportunities In Relation To AI and Islam

Based on the above literature, the following points can be made:

1. If muftis are to navigate the AI landscape and the particular concerns of their Muslim constituents, they cannot simply couch their replies in terms of general Islamic law but need to understand AI in a much more rigorous manner, both with regard to the basic technical issues involved and the deep social implications associated with the field, in formulating their judgments on the issue.
2. Questioners are not comfortable with a fully automated fatāwā. In any case, the physical role of the muftī, as the one who analyses the context of the question, and as the one who will need to confirm the validity of any automated reply, will remain crucial.
3. AI fatāwā have the potential to further encroach upon the traditionally independent role of the muftī and extend further the role of the state. This is because the state, whose resources are typically relied on to initiate a large-scale AI system, can intervene in the type of AI fatāwā that are generated, ensuring that such are not harmful to its interest. This could be the case when a muftī issues a fatwā pointing out an incorrect practice of the head of state or even a verdict regarding the country’s foreign policy.
4. AI systems, such as “smart muftī” and robo-advisor, have the ability to support and fast-track the fatwā provision process. Efficiency and speed is paramount in the age of digital communication and these systems help ensure that fatāwā can respond to users’ needs in a minimal amount of time. But these systems, as noted above, cannot be left to their own devices and require continuous human oversight.
5. The environment created by the surge of AI leads to increasing emphasis on modes of legal reasoning such as collective ijtihād and muftis need to be aware of that. However, such collective ijtihād needs to go beyond a focus on Islamic legal reasoning and speak to other disciplines such as computer science and engineering, broadening the skill-set needed by the muftī while still demonstrating the traditional expertise for which they are respected.
6. While presenting undoubted opportunities for greater efficiency and a ready database, AI fatwā provision faces crucial challenges relating to: understanding the context of a question; expense; national legislation; the opacity and the problem of accountability in relation to an AI system; overreliance on its use leading to a lack of creativity and rigour by the muftī; challenges relating to the training of AI models as “iMuftīs” given the complexity of the knowledge and language involved; and the need for the muftī to be conversant with the technical issues in automated AI fatwā systems.

Given the above, I wish to make some further general observations on the role of the muftī in the light of the challenges and opportunities presented by AI fatwā provision. In short, what follows is my own reflection on, and development of, the points presented in the literature and my own view of the future role of the muftī given these developments.

For me, two challenges particularly impact the muftī whose qualification to provide fatwās derives from being a link in the transmitted chain of Islamic scholarship, and whose interpretation of Islamic law depends on his or her mastery of the Arabic texts. Firstly, as I have shown, the limitations of current AI models in handling the Arabic language pose a challenge, as applying NLP to Arabic for knowledge extraction remains difficult. Secondly, unlike generative AI, pronouncing a fatwā is not merely a mechanical exercise, but rather it is a religious and spiritual obligation that must be sensitive to context. The need to be context-sensitive has been pointed out by almost all the authors above. Every scholar or muftī is accountable and responsible for his or her fatwā and may rightfully be challenged by other scholars. Thus, an explanation of the context of his or her response is vital and, as we have seen, fully automated fatwā systems are, as things stand, not fully attuned to the context of a question. Thus, as is well known among Islamic scholars, when the norms and customs change, the fatwā changes. In this case, context-independent AI may be unaware of these changes and the objectives (maqāṣid) that must be addressed and hence unable to provide a suitable solution. The muftī will be in such a position and should be able to provide a more appropriate response.

A broader challenge is that the issuing of fatāwā has historically transformed from a private activity independent of state control to an increasingly important component of judicial administration. The muftī began as an independent interpreter responding to legal questions within the parameters set forth by an existing school of Islamic legal thought; the position evolved into one that plays a role as a crucial legitimation of the state. In many places, legislation has restricted and constricted their authority. Thus, for example, in the sixteenth century, the Ottoman Empire formally appointed a muftī who served in the office of the Shaykh al-Islam, while in the nineteenth century, the government in Egypt moved to construct the position of an official muftī and an office, known as the Dar al-Ifta to support his work. In the case of AI fatwā provision, the database could very well be under government supervision, and certain views could be excluded or marginalised, as noted by some authors above. This is also tied to another challenge mentioned by authors above: state regulations and partiality to certain Islamic schools of law in specific regions may limit the application of an AI fatwā. Thus, for example, in Islamic finance, the Sharī‘ah authority of a particular country may design or approve the design of a certain financial product and pass a fatwā based on that. The challenge remains when this product has not been approved in another country owing to different financial regulations or different interpretations of Islamic law.

Authors have also noted the fact that disembodied AI cannot fully replicate the sense of community and the sacredness that comes from human connection. In particular, the human ability to express and experience emotion is crucial to religious and cultural practices and interactions. The presence of a physically present imām or muftī holds immense value. Hence, the reliance on such technologies may threaten this vital interaction between communities and their scholars. More particularly, when the question deals with personal matters that require the muftī to understand the questioner’s circumstances and various other human factors and emotions, as is the case with marital issues and some financial situations, AI is difficult to employ, and the human role of the muftī becomes crucial.

The success of a fatwā depends on the skilfulness of the muftī in conceptualising the questioner’s query. The muftī must be discerning, cautious, and alert to the implications of the questions. The reputation and competence of the muftī is considered by the questioner as well as the manner in which he or she makes use of Islamic references. As Latifi pointed out, excessive reliance on AI technology may contribute to a lack of creativity among scholars and an additional disconnect between the scholars and their audience. It may also result in a disconnect between muftis and the texts that they traditionally master. AI may thus result in weakening a scholar’s deductive acumen.

In spite of these challenges, the online globalisation of fatwā allows for more “fatwā shopping” and provides the public with more narratives, presenting an opportunity for more religious pluralism and tolerance. It also reduces bureaucracy in the dissemination of Islamic knowledge. It enables people to obtain answers quickly, thus expediting the process without having to physically consult a muftī. This also makes it cost-effective. It also allows questioners to pose questions of a personal nature without feeling shy or embarrassed. In the case of Islamic finance, speeding up the fatwā process and the advisory services benefits the customers and the institution, and it contributes towards increased revenue. The online option has a far greater reach among Muslims because it transcends borders and other obstacles. The language models could be used to translate fatāwā into multiple languages.

As noted by authors above, AI also stores solutions from previous Islamic legal cases it encounters and can recall the same or a similar situation the next time. This is useful for a muftī who has multiple queries to deal with. For example, the potential smart muftī may be equipped to deal with situations in a particular region or according to a specific school. These smart muftīs can almost fulfil the same role as Sharī‘ah advisors and may be useful in market evaluation, screening products, legal analysis, generating a fatwā database, and using existing information and data to propose new financial products without prior Sharī‘ah rulings.

AI can draw inferences and apply relevant results to the situation. These can then be used by muftis in their fatwā provision. AI chatbots can be trained in languages and the different usages of a word. This could assist the muftī in understanding words in different contexts and societies. It may also be possible to obtain a fatwā “directly” from AI, without the intervention of a muftī if the subject is “straightforward” and merely needs to state whether something is valid or not. An AI robot could fulfil that role as it concerns aspects that may be considered rote learning or involves a mere regurgitation of facts. This is supported by the researchers who maintained that based on collected data an automated question answering chatbot can be built to provide fatāwā for specific questions using deep learning NLP methods. Furthermore, a muftī is advised against issuing a fatwā when feeling hungry or angry, as these emotional states can impede judgment or influence decisions. This difficulty could be obviated using AI or an AI robot because these human characteristics are not applicable.

Finally, given the increasingly widespread use of Large Language Models (LLMs), some remarks on their specific applicability to fatāwā are necessary.

There may be a temptation to skip supervised fatwā platforms in favour of LLMs like ChatGPT owing to their widespread availability. In these situations, care needs to be exercised regarding whether such applications have taken into consideration the different views that exist within the schools of fiqh and the nature and background of the questioner, which currently is not the case. AI systems are not necessarily immune to bias because of their biased datasets, which at times may be perpetuating existing stereotypes. Thus, one can ask as Latifi does: Will it include or exclude the marginalised opinions of scholars? How will AI manage sensitive topics?

Further, the muftī is expected to provide the simplest solution to the questioner and one that promotes less hardship for the questioner. If a fatwā is taken from ChatGPT, how will this system be able to discern a softer opinion from a harsher one? OpenAI uses existing internet datasets to generate its responses. It is a collection of opinions, facts, and knowledge created by everybody and anybody online. It is somewhat like collective intelligence. Because it has access to this huge dataset, it seems omniscient, and its conversation-based coding makes it seem personified. It engages the user and makes them believe that it is the truth. How will an appropriate fatwā be distinguished from an inappropriate one in such an instance?

In a related manner, the Sharī‘ah allows for religious concessions (rukhsah) because of certain circumstances, like a person who is unable to use water having the choice to do the dry ablution (tayammum). Practicing on the concession entails that a specific behaviour or ruling is altered for an individual or a society from a situation that imposed difficulty to one that is easier given the circumstances. Will ChatGPT be able to determine the circumstances that necessitate a concession?

Given the well-known hallucinations associated with LLMs, one must ask how reliable they would be as instruments or mechanisms to provide religious guidance in day-to-day matters, ensuring that people are fulfilling their obligations correctly. Latifi tested this with the same question on a few different occasions and he received different responses each time. This raises the question of reliability. The question of AI hallucination was raised because generative AI can present outputs that appear true or believable but have no real basis in fact.

However, these LLMs may also play a positive role. ChatGPT and similar technologies are advanced databases capable of storing and processing vast amounts of information, analysing it, and subsequently presenting potential solutions. This could be useful to a muftī when he or she needs to use analogy (qiyas) as a legal source. During this process, the muftī needs to know of any precedent or similar situation to the one before him. ChatGPT could assist by analysing thousands of similar scenarios and presenting the muftī with the closest possible scenario. This would save time and expedite a resolution. Thus, regarding Islamic finance, smart muftis can evaluate numerous Sharī‘ah sources, contracts, and rulings across different financial institutions and different schools of fiqh and may then suggest the best possible practice. AI can make informed predictions based on the available information. A muftī must consider the consequences of his fatwā; thus, such predictive capabilities on the part of LLMs could be useful. More broadly, LLMs can also transform data into visuals and diagrams, which may pedagogically aid the muftī in his or her search for a solution.

Conclusion

Many muftis will only reference a few books and often only those common in their respective school of jurisprudence when passing a fatwā. AI enables muftis to explore a vastly larger dataset, allowing them greater access to applicable legal opinions. Thus, I believe that it would be best if the process for issuing fatāwā and how muftis are trained is reviewed in the light of these AI developments. This could involve the establishment of specialised AI-based institutions to train muftis.

However, the human aspect remains central. Given AI hallucinations as well as the extensive training needed to deal with complicated and changing contexts, and to apply different rulings to different individual and social circumstances, the role of the muftī, whose primary role is to be sensitive to such contexts and circumstances, remains critical and will be so for the foreseeable future.

There is a need for competent community-based religious sources and references that are aligned with AI, providing a repository of questions and informed answers, which could be supervised by a panel of scholars or muftis. This panel should also conduct regular workshops or seminars to upskill the muftis given changing developments in AI. Panels of muftis in different countries must cooperate and liaise with international fiqh academies in compiling contemporary fatāwā and coin new, appropriate Islamic legal terminology given unfolding technological developments. AI could naturally assist and expedite the process by gathering or serving as a repository for all the different fatāwā on a topic.

Ultimately, AI processes rely on the individuals who develop them, which means they can be both used and misused. Every muftī must be familiar with the developments and changing circumstances that firstly impact the society he or she lives in and thereafter the global impact. The role of the muftī should evolve to incorporate technology as a valuable tool rather than something that is imposed from the outside. The muftī must understand how AI works along with being able to distinguish between strong and weak AI. This would make them more suitable to pronounce a fatwā and they would be better placed with experts in the field to develop appropriate datasets. Hybrid models that employ AI as preliminary filters for fatwā queries, supplemented by human review and personalised feedback, may offer a promising way forward. The partnership between the two could enhance and enrich the fatwā provision process, making it more efficient and accessible.

References

• Abdussalam, I. D. n.d. Qawaid al-Ahkam fi Masalih al-Anam. Dar al-Kutub al-Ilmiyah.
• Al-Asqalani A. 1987. Fath al-Bari bi Sharh Sahih al-Bukhari. Dar al-Rayyant.
• Al-Ayni, M. F. 2018. Risalat fi Adab al-Mufti. Nashriyat Waqf al-Diyanat al-Turki.
• Al-Farfur, W. 1999. al-Mudhhib fi Usul al-Madhhab ala al-Muntakhab. Maktabat Dar al-Farfur.
• Al-Farfur, W. 2010. al-Fatwa, Arkanuha, Dawabituha, Shurutuha, Adabuha wa Manahij al-Ifta. Dar al-Farfur.
• Al-Ghazali, M. 1971. Shifa al-Ghalil fi Bayan al-Shubah wa al-Mukhayl wa Masalik al-Ta’lil. Matba’at al-Irshad.
• Al-Jawziyah, I. Q. 1998. I’lam al-Muwaqqi’in an Rabb al-Alamin. Dar al-Kitab al-Arabi.
• Al-Khin, M. 1993. Athar al-Adillat al-Mukhtalaf fiha fi al-Fiqh al-Islami. Dar al-Qalam.
• Al-Qarahdaghi, A. M. 2018. al-Ijtihad wa al-Fatwā. Dar al-Nida.
• Al-Qarafi, S. D. n.d. al-Furuq. Dar al-Marifah.
• Al-Qarafi, S. D. 1995. al-Ihkam fi Tamyiz al-Fatawa an al-Ahkam wa Tasarrufat al-Qadi wa al-Imam. Maktab al-Matbu’at al-Islamiyah.
• Al-Nawawi, Y. 1988. Adab al-Fatwa wa al-Mufti wa al-Mustafti. Dar al-Fikr.
• Al-Raysuni, Q. 2014. Sina’at al-Fatwā fi al-Qadaya al-Mu’asirah. Dar Ibn Hazm.
• Almansur L. 2024. “Al Fatwā án tariq al Dhaka al Istina’I bayn al iqdam wa al ihjam.” Magazine of the College of Islamic Studies and Arabic 9 (1): 451–474.
• AlShithisri, A. A. 2023. “A Study of Legal and Fiqhi Impact of AI on Issuing Fatwās.” Kurdish Studies 12 (1): 2504–2511. https://doi.org/10.58262/ks.v12i1.174
• Al-Suyuti, J. n.d. al-Ashbah wa al-Naza’ir. Dar al-Fijr.
• Al-Zarkashi, M. 1992. al-Bahr al-Muhit. Ministry of Endowments and Islamic Affairs, Kuwait.
• Brown, N. 1999. “Reviewed Work: Defining Islam for the Egyptian State: Muftīs and Fatwās of the Dar al-Ifta, Jakob Skovgaard-Petersen.” International Journal of Middle East Studies 31 (2): 282–283. https://doi.org/10.1017/S0020743800054106
• Dawudi, S. 2010. Qawa’id Usul al-Fiqh wa Tatbiqatuha. Dar al-Asimah.
• Elhalwany, I., A. Mohammed, K. Wassif, et al. 2015. “Using Textual Case-Based Reasoning in Intelligent Fataw Systems.” The International Arab Journal of Information Technology 12 (2): 503–509.
• Ghaly M. 2023. “What Makes Work ‘Good’ in the Age of Artificial Intelligence (AI)? Islamic Perspectives on AI-Mediated Work Ethics.” The Journal of Ethics 28: 429–453. https://doi.org/10.1007/s10892-023-09456-3
• Hallaq, W. B. 2002. “A Prelude to Ottoman Reform: Ibn ‘Abidin on Custom and Legal Change.” In Histories of the Modern Middle East: New Directions, edited by I. Gershoni, Y. Hakan Erdem, and Ursula Woköck, 37–61. Lynne Rienner Publishers. https://doi.org/10.1515/9781588269867-005
• Hassan, A. S. 2015. “The Muftī–Mustafti Approach to Religious Rehabilitation.” Counter Terrorist Trends and Analysis 7 (3): 14–20.
• Hasyim, M. S. 2020. “A New Islamic Knowledge Production and Fatwā Rulings: How Indonesia’s Young Muslim Scholars Interact with Online Sources?” Journal of Indonesian Islam 14 (2): 499–518. https://doi.org/10.15642/JIIS.2020.14.2.499-518
• Ismail, A. M., and A. S. Baseeruddin. 2022. “Moderation in Fatwā and Ijtihad.” AHKAM: Jurnal Ilmu Syariah 22 (1). https://doi.org/10.15408/ajis.v22i1.22299
• Jackson, S. 1992. “The Second Education of the Mufti: Notes on Shihab al-Din al-Qarafi’s tips to the Jurisconsult.” The Muslim World 82 (3–4): 201–217. https://doi.org/10.1111/j.1478-1913.1992.tb03553.x
• Kausar S., A. R. Leghari, and A. S. Soomro. 2024. “Analysis of the Islamic Law and its Compatibility with Artificial Intelligence.” Annals of Human and Social Science 5 (1): 99–114. https://doi.org/10.35484/ahss.2024(5-I)10
• Latifi, H. 2024. “Challenges of Using Artificial Intelligence in the Process of Shi’i Ijtihad.” Religions 15 (5): 541. https://doi.org/10.3390/rel15050541
• Mahmood, S. F., et al. 2023. “Fiqh Robotic for Artificial Intelligence in Humanoids used for therapy, services and other social activities: An Integration of AI and Maqasid Shariah.” Journal of Fatwā Management and Research 28 (2): 527.
• Malik, S. A. 2024. “Artificial Intelligence and Islamic Thought: Two Distinctive Challenges.” Journal of Islamic and Muslim Studies 8 (2): 108–115. https://doi.org/10.2979/jims.00020
• Masud, M. K., B. Messick, and D. Powers. 1996. “Muftis, Fatwas and Islamic Legal Interpretation.” In Islamic Legal Interpretation, Muftīs and Their Fatwās, edited by M. K. Masud, B. Messick, and D. Powers, 3–32. Harvard University Press.
• Mohd, W., et al. 2018. “The Application of Technology in the Dissemination of Fatwā: A Study on Religious Institutions in Malaysia.” International Journal of Civil Engineering and Technology 9 (7).
• Munshi A. A., et al. 2021. “Towards an Automated Islamic Fatwā System: Survey, Dataset and Benchmarks.” International Journal of Computer Science and Mobile Computing 10 (4): 118–131. https://doi.org/10.47760/ijcsmc.2021.v10i04.017
• Munshi A. A., et al. 2022. “Automated Islamic Jurisprudence Legal Opinions Generation Using Artificial Intelligence.” Pertanika Journal of Science & Technology 30 (2): 1135–1156. https://doi.org/10.47836/pjst.30.2.16
• Nord, I. 2023. “The Appreciation of ‘Things’ in an Ambiguous World: On AI & Religion.” In Thinking Tools on AI, Religion and Culture, 43–46.
• Polat, A., et al. 2022. “An Inquiry into the Application of Artificial Intelligence on Fatwā.” In Digital Transformation in Islamic Finance, 273–286. https://doi.org/10.4324/9781003262169-18
• Rhamadhi L. 2024. “Istithmar Taqniyat al Dhaka al Istina’i fi Sina’at al Fatwā li al Mu’assasat al Maliyah al Islamiyah.” Research and Applications in Islamic Finance 8 (1): 55–74.
• Sa’ad, Adam, et al. 2019. “Robo-Advisory for Islamic Financial Institutions: Shariah and Regulatory Issues.” European Journal of Islamic Finance.
• Salhein, Y. 2018. “State–Society Relations in Digital Fatwās: A Study of Islamweb in Qatar.” MA thesis.
• Shuqayrat, A. S. 2002. Tarikh Mu’assasah Shuyukh al-Islam fi al-Ahad al-Uthmani. al-Maktabah al-Wataniyah.
• Singer J. 2021. “Fatwā’s from Islamicweb.net on Robotics and Artificial Intelligence.” In Artificial Intelligence in the Gulf, 279–303. https://doi.org/10.1007/978-981-16-0771-4_12
• Tsourlaki, S. 2022. “Artificial Intelligence and Fatwā Issuance: A Case Study of Dubai and Egypt.” Islamic Inquiries 1 (2): 107–125.
• Tsuria, R. 2023. “AI and Religion: The Voice of God in Everyday Life.” In Thinking Tools on AI, Religion and Culture, 30–33.
• Wahid, S. H., and A. Kususiyanah. 2020. “The Relevance of the Usage of Artificial Intelligence and Machine Learning in Legal Analysis.” In Proceedings of ICORSH 2020.
• Yahya, I., and Junaidi, A. A. 2021. “Authority of Fatwā in the Digital Era.” Webology 18 (2).
• Yucel, S., and A. Ismail. 2021. “Are Online Fatwas Credible?” Religions 12: 902. https://doi.org/10.3390/rel12100902