E-Prosiding Persidangan Serantau Manasik Haji 2025 (MANASIK2025)

MANASIK2025 Putrajaya, Malaysia, 7-8 Oktober 2025 634 logs) adapted from AI accountability frameworks (Raji et al., 2020) and tailored to fatwa risk profiles (invalidating rituals, incurring dam ). 4. Corpus engineering for Hajj scenarios. TH’s Muzakarah corpus is rich but under -structured. The article proposes a schema (ritual phase, actor state, constraint, ruling, remedy) to enable faithful retrieval and scenario matching (Lembaga Tabung Haji, n.d.). 5. Human-in-the-loop design. Current systems over-index either on static FAQs or unconstrained generation. The article specifies escalation pathways to muftis and TH counselors, recognizing institutional realities observed across Malaysia. (e-Fatwa Mufti WP, n.d). Toward a Sharīʿah - compliant AI design for Hajj/ʿUmrah Bringing these strands together, the most persuasive direction is a constrained, auditable, and jurisdiction-aware assistant: (i) built on a curated, versioned corpus (state fatāwā, national Muzakarah, TH resolutions), (ii) delivered via RAG with inline citations to the exact sighah , (iii) governed by institutional audits and incident response, and (iv) instrumented for escalation when confidence or coverage is inadequate. This approach aligns with institutional digitization trends in Malaysia, respects state authority boundaries, and applies the broader AI reliability literature to a religious context without automating ijtihād . The contribution is not merely to “use AI,” but to encode governance into the system — so that every answer carries its chain of authority as visibly as its text. Theoretical Framework The development of a Sharīʿah -compliant framework for AI-driven fatwa automation in the Malaysian Hajj and ʿUmrah context requires grounding in multiple theoretical perspectives that account for religious authority, technology adoption, and ethical governance. This section outlines four interrelated frameworks — uṣūl al -fiqh and fatwa theory , Maqāṣid al - Sharīʿah , Technology Acceptance and Trust Models , and AI Governance and Accountability Frameworks — and demonstrates how each supports the study’s objectives. Uṣūl al -Fiqh and Fatwa Theory The first theoretical anchor is the Islamic jurisprudential framework of uṣūl al -fiqh , which defines the principles and conditions for valid fatwa issuance. Classical scholars such as al- Ghazālī and al - Shāṭibī stressed that fatwa requires competence, contextualization, and evidentiary reasoning ( istidlāl ) based on the Qurʾān, Sunnah, consensus ( ijmāʿ ), and analogical reasoning ( qiyās ) (Kamali, 2003). In Malaysia, these theoretical underpinnings are institutionalized within legal-procedural rules that vest authority in the State Mufti Committees, subject to gazetting by the ruler. By embedding uṣūl al -fiqh principles, the study ensures that AI-driven systems are not treated as independent jurists but as supportive tools. Fatwa theory also stresses the importance of accountability ( masʾūliyyah ) and the differentiation between binding rulings and advisory opinions. This directly informs the study’s objective of conceptualizing AI systems that are jurisdiction-aware , tethered to state-level authority, and limited to authenticated corpora rather than open-ended ijtihād . Maqāṣid al - Sharīʿah The second framework is Maqāṣid al - Sharīʿah (higher objectives of Islamic law), which provides an ethical and purposive lens for evaluating technology in the religious domain. Traditionally articulated by al- Shāṭibī and refined in contemporary scholarship (Auda, 2008), the maqāṣid emphasize the protection of religion ( dīn ), life ( nafs ), intellect ( ʿaql ), dignity ( ʿird ), and wealth ( māl ). This framework is highly relevant because the risks of AI-driven fatwa automation-misguiding pilgrims, undermining mufti authority, or exposing users to invalid rituals-directly implicate the protection of dīn