The newest Shariah question there is. We advise builders of AI products for Muslim users on what a system may say, what it must not, and who is accountable when it is wrong.
The problem
An AI built to guide Muslims produces confident, fluent, individualised statements about what a person may and may not do, at a scale no scholar could review, to users who are not positioned to judge whether the answer is sound. Fluency and soundness are different things, and the person receiving an answer cannot tell them apart.
The questions that follow are new and largely unsettled. May a system that is not a person issue what functions as a fatwa? What is the status of an answer that is plausible, well expressed and wrong? Which school does it follow, and does it tell the user? What must it refuse outright? How does a difficult question, one where scholars differ, or where the ruling turns on facts the system does not have, get escalated to a human rather than answered?
These are Shariah problems that happen to be implemented in software rather than engineering problems with Shariah implications, and they have to be answered in the product design rather than in a disclaimer.
Can it issue a fatwa?
Our position: no, and the design must reflect that
Is the answer sound?
Fluent and correct are not the same thing
Does it disclose its school?
And tell the user what they are receiving
Does it escalate?
To a qualified human when it should
What we do
What the system answers, what it declines, and where it must escalate to a qualified human. The refusal set matters more than the answer set.
Which sources the system draws on, which school it follows, and how the user is told what they are receiving.
Where the system should express uncertainty rather than confidence, ihtiyat built into the product rather than bolted on afterwards.
Periodic review of the responses the system generates, with anything non-compliant filtered out. In practice this runs in two tiers: moderators raise primary flags on what the system produces, and our Shariah team reviews what is escalated while also sampling independently.
Who is answerable for what the system says, what human oversight the product requires, and how that oversight is evidenced.
What the user is told about what this is, what it is not, and when to ask a person instead.
Certification under a verifiable reference, and continuing review as the model, the data and the product change, which for an AI product is constantly.
Why ADL
Our team pairs certified Muftis holding AAOIFI qualifications with people who understand code, systems and product architecture. We hold Malaysia Digital Status from the Malaysian Digital Economy Corporation, a technology credential, held by a Shariah advisory firm. Where there is an app, we go into it. Where there is a core banking system, we review the configuration.
Registered Shariah Adviser with the Securities Commission Malaysia and with Labuan Financial Services Authority.
We work primarily against AAOIFI standards and reconcile with the applicable local regime. An opinion grounded in AAOIFI travels across borders.
Beyond applying Islamic finance standards, our team has been commissioned by a standard-setting body: research towards a governance standard, participation in the drafting of a preliminary standard, and a series of training assignments.
Auditing against another adviser's pronouncement is ordinary work for us, as is being audited by another firm. Independence is the point of the exercise.
We work in jurisdictions with mature Islamic finance regulation and in markets with none at all, where the structure has to satisfy Shariah while operating entirely within a conventional rulebook.
Best Shariah Advisory in Islamic Asset Management, presented in Jeddah in February 2026.
Members of our team hold Shariah board and committee seats across the institutions we serve, so our advisory work is informed by governance experience, not only by external review.
How we hold ourselves
Our measure of a good year is not only revenue. It is whether we helped one more business get to halal. That is why we will take a call from a founder at ideation stage, and why our pricing bends to what a client can actually carry. We would rather a small platform get its structure right at the start than be priced out and get it wrong at scale.
A business built on an Islamic label still has to be a good business. Sound fundamentals, capable people, honest disclosure, and a high standard of compliance with local regulation, because that is what protects the customer and the investor. Shariah compliance sits on top of that foundation. It is not a substitute for it, and we will say so if we see it being treated as one.
The Lifecycle
Compliance is a state you maintain rather than a certificate you obtain, and the audit cycle is how you maintain it.
We understand the product, the model and the user base before we quote.
Scope, timeline and fee agreed, agreement signed.
The deep stage: scope and refusal design, source and attribution, and how uncertainty is handled.
Findings returned and worked through with your product and engineering teams; the refusal and escalation design is adjusted.
Issued by a certified Mufti.
Issued under a reference any counterparty can verify independently.
For an AI product the cycle runs continuously: outputs are sampled on an ongoing basis, not reviewed once and left, because the model, the data and the product change constantly.
We commit to two to four weeks for a first pronouncement. In practice it often runs longer, and in our experience the reason is the round trip rather than the review. We raise amendments, and your product and engineering teams need time to work through them. We would rather set that expectation now than surprise you in week three.
Engagements are scoped and priced individually. We are advising the development of a Shariah-compliant conversational AI for Muslim users. One size does not fit all. Tell us what you are building and we will tell you what it takes.
Evidence
An AI company (Southeast Asia)We are advising the development of a Shariah-compliant conversational AI, a companion product intended to give Shariah-grounded guidance to Muslim users. An ongoing engagement.
Questions
Our working position is that it cannot, and that a product which behaves as though it can is the central risk to design against. A system can convey established rulings, explain reasoning, and point a user to sources. Ifta is the act of a qualified person exercising judgment on a particular situation, and it carries an accountability that software cannot bear. Where exactly that line falls in a live product is most of what the design work is about.
Different in kind from a wrong answer in most other domains. A user may act on it in worship, in marriage, in inheritance or in money, and may never discover it was wrong. That asymmetry is why refusal and escalation design matters here more than any accuracy benchmark.
The outputs. Sampling what the system says is the only way to know what users receive. It is the same principle that has us reading smart contract code rather than the whitepaper, and core banking configuration rather than the product guideline.
Partly, and the questions shift. There the issue is usually whether an AI-driven recommendation, pricing or underwriting decision can produce an outcome the Shariah structure does not support, and whether anyone would notice.
Our certification confirms Shariah compliance. It is not an assessment of whether a business is viable, whether its financials are sound, or whether the people running it can execute. Look at both questions, and do your own due diligence on the second. Every ADL certificate carries a reference you can check independently.
Book a consultation, whether you are designing the refusal set now or reviewing what your product has already shipped.