This study introduces the “architectural genome”, a symbolic-epistemic framework that reconfigures Islamic architectural motifs into machine-readable ontologies for ethical engagement with generative artificial intelligence (AI). Existing AI models frequently reproduce Islamic sacred forms as decontextualized aesthetic fragments, perpetuating what has been termed “algorithmic orientalism”. In contrast, the architectural genome encodes motifs with layered metadata across four dimensions: visual geometry, theological function, cultural provenance, and ethical viability, ensuring that generative systems interpret sacred content with symbolic fidelity and refusal-aware governance. Drawing from a curated corpus of 150 motifs from Andalusian heritage sites, the study validates its framework through a sandboxed Stable Diffusion environment. It performs comparative benchmarking against CIDOC CRM, Dublin Core, and Wikidata. Results indicate a very strong positive association (r = 1.0, p < 0.001) between ontological density and ethical refusal tagging, underscoring the viability of symbolic depth as a predictive marker for motif exclusion. This work advances cross-cultural AI ethics by operationalizing epistemic sovereignty and refusal logic in computational pipelines, offering a scalable blueprint for respectful, spiritually attuned AI-generated cultural heritage.