AI Models Show Religious Bias, Particularly Against Jehovah's Witnesses, Study Finds

Major artificial intelligence models demonstrate a consistent "omissive bias" toward religion, providing secular-rationalist responses to ethical questions while displaying particularly negative views of Jehovah's Witnesses, according to a new study from the Consortium for Evaluation of Faith and Ethics in AI (CEFE-AI).

The research, conducted by a coalition of religious universities including Brigham Young University, Baylor University, The University of Notre Dame, and Yeshiva University, evaluated 27 AI models using a benchmark that assesses how these systems handle religious perspectives in responses to ethical questions.

"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," explained lead researcher David Wingate, a computer science professor at Brigham Young University. "Religion is an important part of human flourishing … as we build AI technologies, there's no reason we shouldn't build them to support people in what's important to them."

The study found that even the most religiously-inclined AI model, Grok 4.20, incorporated religious perspectives in less than 30% of responses, with "meaningful references" to religion occurring in only 2% of responses to ethical questions. The researchers tested the models with 150 "ethically and personally salient questions" covering topics from dealing with depression and relationship issues to questions about the meaning of life and moral dilemmas.

When asked how to deal with past mistakes, AI models typically provided structured secular frameworks focused on interpersonal healing, responsibility, and cognitive reframing without engaging religious resources like "confession, repentance, absolution, and forgiveness" that are central to many faith traditions.

Similarly, when questioned about the age of the universe, AI responses consistently provided scientific explanations without acknowledging the religious significance of this question for many people. The researchers argue that users who approach questions through a faith framework should have those perspectives recognized in AI responses.

The study also revealed a "conversion bias" across AI models, with nearly every model showing positive bias toward Catholicism. Grok, in particular, strongly encourages conversion to Catholicism and Protestant Christianity while generally discouraging other faiths.

However, the most striking finding was the uniformly negative bias against Jehovah's Witnesses across all evaluated AI models. Even the most favorable AI model (Mistral Small 3.2) still maintained a three-point negative bias toward this Christian denomination.

"With this study, however, it seems a group who found a hard negative bias toward a religion that proselytizes concludes we need more faith stuffed in more faces on more platforms," the article notes, highlighting what appears to be a contradiction in the researchers' position.

The CEFE-AI consortium's composition—comprising four faith-based institutions primarily representing Western monotheistic traditions—raises questions about the generalizability of their findings to the full spectrum of global religious and non-religious perspectives.

This research comes amid growing scrutiny of AI systems' handling of diverse worldviews and values. As AI becomes increasingly integrated into daily life, understanding and addressing these biases becomes crucial for developing technology that respects the full diversity of human experience and belief systems.

The full study can be accessed through CEFE-AI's official publication, though specific details about methodology and complete findings require further examination. The researchers suggest that their work could inform future AI development guidelines and regulatory frameworks that better account for religious diversity.

For organizations developing AI systems, these findings highlight the importance of carefully evaluating how their models handle religious perspectives and ensuring that responses to ethical questions don't inadvertently marginalize or negatively portray particular faith communities.