Transparent Artificial Intelligence in Theology Classrooms: Ethical and Pedagogical Insights of the Apertus Model for Religious Education

This study investigates how openness in AI design influences theological learning in religious education (RE). Drawing on Zizioulas's theology of personhood and Floridi's information ethics and aligned with the Greek upper‐secondary RE curriculum, an open‐source multilingual model (Apertus) is compared with a proprietary system (GPT‐5). Fifty curriculum‐based prompts generated paired 200‐word responses, evaluated by three theologian‐educators on ethical transparency, inclusivity, doctrinal responsibility, and stylistic fluency. Evaluations were conducted by three theologian‐educators with professional experience in university‐level religious studies and ethics, ensuring alignment with contemporary higher‐education discourse. Apertus scored significantly higher in the first three dimensions (Cohen's d = 1.28–3.46), with strong inter‐rater agreement ( κ = 0.85). Qualitative findings confirmed that Apertus grounded dignity in the image of God ( imago Dei ), treated interfaith dialogue as a theological vocation, and linked digital ethics to stewardship with traceable sources, while GPT‐5 produced fluent but secularized answers. It is concluded that openness and configurability determine the ethical and theological reliability of artificial intelligence in RE, advancing dialogue, freedom of belief, and human dignity.