Researchers from a consortium of universities, including Baylor, Brigham Young, Notre Dame, and Yeshiva, released the AllFaith Benchmark on Tuesday, revealing that major AI models systematically favor certain religions. The study analyzed 3,640 responses from 20 models and found Catholicism received a 61% positive rating, while Jehovah's Witnesses got only 3%. Agnosticism scored highest at 71%. Grok 4.20 showed the strongest bias, favoring Catholicism (69%) and Evangelical Protestant (51%). The findings come a day after Pope Leo XIV issued an encyclical warning that AI absorbs creators' values. Only 0.2% of AI bias papers address religion.

Key facts

• AllFaith Benchmark tested 20 AI models on 3,640 religious queries
• Catholicism favored at 61% positive; Jehovah's Witnesses at 3%
• Grok 4.20 showed strongest religious bias overall
• Only 0.2% of AI bias studies examine religious bias
• Pope Leo XIV warned AI reflects creators' values

KeyAudit data perspective