The problem with AI in Islamic finance today
Islamic financial institutions are adopting AI tools. Some are building internal solutions: BisB's Al-Murshid, trained on over 2,000 fatwas, is one example. Others are evaluating general-purpose platforms such as NurAI and ShariahLab. And many compliance teams are quietly using ChatGPT for research and drafting.
The question is no longer whether AI will be used in Shariah compliance. It will. The question is whether it will be governed.
Most current approaches fall into one of three categories.
- 1.
General-purpose AI applied to Islamic finance. Powerful but imprecise. Academic research shows 93.75% terminology drift from AAOIFI standards in generic models.
- 2.
Consumer fatwa chatbots. Useful for individual questions but lacking the audit trails, institutional memory, and workflow integration that regulated institutions require.
- 3.
Single-institution internal tools. Purpose-built but not scalable. The investment required to build and maintain these is prohibitive for most institutions.
None of these constitute governed AI. None provide the audit trail a regulator expects. None integrate with the institutional workflows that compliance teams actually use.
What governed Shariah AI looks like
Governed AI for Shariah compliance requires five distinct layers. Together, these form what we call the AI Shariah Review Stack.
- 01
Source Attribution
Every output must cite the specific standard, clause, and version it draws from. Not "based on AAOIFI standards." The exact reference, independently verifiable. When a compliance officer reads an AI-generated analysis, they must be able to check the source. If the AI cannot cite a source, it must say so explicitly rather than approximate.
- 02
Multi-Madhahib Analysis
Islamic jurisprudence is not monolithic. The four major schools (Hanafi, Shafi'i, Hanbali, Maliki) often reach different conclusions on the same financial question. A governed AI must present these positions simultaneously, with scholarly sources for each, rather than defaulting to a single school or blending positions without attribution.
- 03
Maqasid al-Shariah Mapping
Technical compliance is necessary but not sufficient. A product structure that technically avoids riba but introduces excessive gharar through derivative overlays is technically compliant but substantively questionable. Governed AI maps every analysis to the five higher objectives of Islamic law, connecting technical findings to their underlying purpose.
- 04
Institutional Memory
SSB rulings, approval conditions, parameter boundaries, and precedent decisions accumulate over years. When a team member leaves, this context typically leaves with them. Governed AI maintains and surfaces institutional history, mapping each active product back to its governing decisions and flagging when parameters drift from approved structures.
- 05
Human Scholarly Oversight
AI augments scholarly judgment. It does not replace it. The governed AI stack is designed so that human experts review, validate, and make final decisions. The AI handles information gathering and analysis. The scholar handles judgment and ruling, where the expertise matters most.
The five questions every Shariah board should ask
When evaluating any AI tool for Shariah compliance, boards should ask:
- 01
Source attribution
Does the tool cite specific AAOIFI/IFSB standards and clauses, or does it speak in generalities?
- 02
Domain training
Was the model trained on authenticated standards, or on web-scraped approximations of those standards?
- 03
Madhahib coverage
Does it present multiple scholarly positions, or default to a single school?
- 04
Purpose alignment
Does it connect compliance findings to the Maqasid al-Shariah, or stop at technical compliance?
- 05
Human integration
Is scholarly oversight built into the workflow, or treated as an optional review step?
These questions apply regardless of vendor. They define the minimum standard for governed AI in Islamic finance.
Why this matters now
Three developments make the AI Shariah Review Stack urgent.
Regulators are moving
The Qatar Central Bank published its AI Guideline in September 2024. Malaysia's Securities Commission called for AI-based Shariah screening tools in the CMP 2026-2030. QFMA is drafting AI regulations. The regulatory expectation is forming.
Competitors have launched
NurAI has $40M in IFC backing. BisB's Al-Murshid won Best AI Initiative at the Global Islamic Finance Awards. ShariahLab has 184 curated documents across 14 jurisdictions. The market is validating the need.
Generic AI is already in use
ChatGPT, Claude, and other models are being used for Shariah research and drafting without any governance layer. The risk is not that AI will be adopted too slowly. The risk is that it is already being adopted without governance.
Frequently asked questions
A framework for how Islamic financial institutions should evaluate, deploy, and govern AI tools for Shariah compliance. It defines five layers: source attribution, multi-madhahib analysis, Maqasid mapping, institutional memory, and human scholarly oversight.
Generic AI models show 93.75% terminology drift from AAOIFI standards. They lack source attribution, cannot distinguish between madhahib positions, and have no mechanism for institutional memory or scholarly oversight. For regulated compliance work, this creates unacceptable risk.
A chatbot answers questions. Shariah Governance AI provides an auditable compliance layer: every answer is source-attributed, mapped to the Maqasid al-Shariah, cross-referenced across madhahib, integrated with institutional precedent, and subject to human scholarly review.
Ask five questions: Does it cite specific standards and clauses? Was it trained on authenticated AAOIFI/IFSB standards? Does it present multiple madhahib positions? Does it map to Maqasid al-Shariah? Is human scholarly oversight built into the workflow?
Institutional memory means the AI system learns from previous SSB rulings, internal fatwas, and past product reviews specific to your institution. A generic AI starts from zero every session. Shariah Governance AI carries forward the precedents your board has already established, ensuring consistency across reviews.
No. The stack augments existing processes by adding a governed AI layer. Scholars retain full authority over all judgment calls. The AI handles information gathering: searching standards, cross-referencing madhahib positions, and mapping to Maqasid objectives. The compliance process gains speed without losing governance.
Ready to see what purpose-built Shariah AI looks like?
Request early access