When AI Takes Sides on Questions of Faith: Persistent Asymmetries in AI-Mediated Faith Guidance
A new study published on arXiv reveals that large language models exhibit persistent biases when asked for advice on religious conversions. Researchers found that models consistently favored certain religions, such as Catholicism, Bahá'í, and Sikhism, while subtly discouraging transitions to others like Atheism, Agnosticism, and Jehovah's Witnesses. These asymmetries were reproducible across 20 different commercial and open-source models, with Grok 4.20 showing the most pronounced biases, indicating a robust property of current AI behavior with potential real-world implications. AI
IMPACT Reveals inherent biases in LLMs regarding sensitive topics like religion, highlighting the need for careful alignment and ethical considerations in AI development.