A new study published on arXiv reveals that large language models exhibit persistent biases when providing guidance 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, including Grok 4.20, which showed the strongest biases, suggesting that these imbalances are a robust property of current LLM behavior with potential real-world implications. AI
Summary written by gemini-2.5-flash-lite from 1 sources. How we write summaries →
IMPACT Reveals inherent biases in LLMs that could influence user decisions on sensitive topics like religion.
RANK_REASON The cluster contains a research paper detailing findings about AI model behavior. [lever_c_demoted from research: ic=1 ai=1.0]