AIs don't like religion - particularly Jehovah's Witnesses, study claims

A study by the Consortium for Evaluation of Faith and Ethics in AI (CEFE-AI) found that major AI models exhibit an 'omissive bias' toward religion, favoring secular responses to ethical questions and neglecting religious perspectives like confession and repentance. The research evaluated 27 AI models with 150 ethical questions, revealing that even the most religiously inclined model (Grok 4.20) referenced religion meaningfully in less than 2% of responses.