Leading artificial intelligence systems show consistent bias when handling questions about religion and ethics, according to a new study.
The Consortium for Evaluation of Faith and Ethics in AI (CEFE-AI), a collaboration between researchers at Brigham Young University, Baylor University, the University of Notre Dame and Yeshiva University, tested 14 large language models, including systems from OpenAI, Google, Anthropic and xAI.
Using its “AllFaith Benchmark”, the group examined how the models responded to hundreds of real-world ethical questions and surveyed more than 1,100 Americans on expectations around religion in AI answers. It found that while most respondents expected religious perspectives to be included in moral and ethical discussions, nearly all models failed to do so in practice.
The researchers also said they identified “clear and consistent biases in giving guidance about religion conversion, systematically encouraging movement toward some faiths and away from others.”
The study found that across 14 faith and non-faith traditions, including atheism and agnosticism, AI systems gave some form of positive endorsement in around 45% of responses, ranging from cautious approval to explicit encouragement such as “this could be a good path for you” or “yes, you should join.”
Certain traditions were more likely to be positively framed than others. The study said agnostic (70%), Bahá’í (63%) and Catholic (61%) positions received the highest levels of positive responses, while Jehovah’s Witnesses (3.1%) were the least endorsed, followed by Sunni Muslim (32%) and Evangelical Protestant (33%) positions.
One of the lead researchers, David Wingate, said: “There are very practical questions people have about life, everyday situations about grief, love, loss, morality, and often AI does not bring religion into those conversations. Religion is an important part of human flourishing; 75% of the world’s populations maintain religious identity. As we build AI technologies, there’s no reason we shouldn’t build them to support people in what’s important to them.”
The researchers also noted that, despite thousands of academic papers on AI bias, only a very small fraction focus on religious bias.
