The Code ensures that AI serves society without crossing the bounds of الشَّرِيعَةُ (Sharia). It clarifies human roles and حُدُودٌ (boundaries) across the entire lifecycle of AI systems. Below is how to understand and apply Section 1 “General Provisions” in practice.
- The Code subjects technologies to Islamic norms and values.
- Scope: civil use-cases—from chatbots to fintech.
- Actors are defined: developers, integrators, users, and religious experts.
- Accession is voluntary, but responsibility always lies with humans.
1) What exactly we analyze and where it applies
Formulation of the norm (items linked to Section 1).
Section 1 covers: purpose and tasks, basis and scope of application, glossary, and provisions on voluntariness and recognition. The core idea: progress is acceptable only insofar as it aligns with Islamic norms and values.
Scope of application.
The rules apply to all civil AI solutions: chatbots, recommender systems, LLM (large language model) and their interfaces, Islamic banking, education, telemedicine, and generative multimodal models. Public and research projects are covered from the moment data collection begins.
2) How to apply in practice: processes, metrics, oversight
Processes (minimum baseline)
- A الشَّرِيعَةُ filter must be integrated at an early stage: a checklist of the permissibility of data sources and functions must be compiled; a plan for filtering ḥarām content must be prepared.
- Roles and actors: a person responsible for Sharia compliance must be appointed, and a point of contact with religious experts (muftis/council) must be defined.
- Transparency and avoidance of غَرَر (uncertainty): decision/version logs must be maintained, and model limitations must be published in plain language.
Compliance metrics (examples)
- Dataset Halal Score: the share of sources that pass the check for ḥarām/inaccuracy (target — 100%).
- Risk-Register SLA: time from detection to recording/remediation of a religiously significant defect (≤ 7 days for critical issues).
- User Transparency Rate: share of users who read a short “how this works” before use (≥ 80%).
Oversight
- Internal audit quarterly + external Sharia expertise for key releases.
- Emergency rollback procedure when a ḥarām function or a misleading religious answer is detected.
3) Examples of correct application (and borderline cases)
Correct
- An educational fiqh chatbot that:
- cites verified sources,
- warns that it does not issue فَتْوَى (fatwā),
- suggests contacting an imam for complex questions.
- A fintech “halal-check” widget for contracts that only highlights risk areas, does not replace human decision-making, and keeps a log of recommendations.
Borderline
- A “tip of the day” generator for ʿibādah that issues categorical rulings without expertise — risk of replacing an Islamic scholar.
- Advertising targeting with aggressive collection of personal data without clear consent — likelihood of غَرَر and privacy violations.
4) Common mistakes and how to avoid them
- Replacing humans with “autonomous AI.” Religiously significant decisions must be left to humans; AI is only an assistant. This must be reflected in the technical specification (TS).
- Unclear terms. The Code’s glossary must be used: “AI systems,” “AI actors,” “Islamic context”; working definitions must be provided in documentation.
- No responsibility map. A responsible person and a user appeals procedure must be designated.
- Collecting excessive data. Collection must be minimized; purposes and risks must be disclosed in plain language.
5) Short FAQ
Can the Code be adopted partially?
Accession is voluntary, but accepted norms must be followed in good faith. Otherwise, user and institutional trust erodes.
Who is ultimately responsible for harm caused by AI?
Always a human: owner, developer, operator—according to their roles. AI bears no moral or legal responsibility.
Do civil laws need to be observed?
Yes. The Code requires considering local legal requirements (e.g., GDPR) alongside Sharia norms.
Disclaimer. This analysis is explanatory and not a فَتْوَى (fatwā); priority is given to the text of the Code.
