< formatted article >

AI and Religion: The Silent Bias in Big Tech’s Most Powerful Models

A groundbreaking study has uncovered a startling truth: when it comes to religion, AI is eerily quiet—and when it does speak, it has a noticeable bias.

Researchers from four universities developed a test called AllFaith to examine how major AI models—including those from Google, OpenAI, and Anthropic—discuss religion. The experiment posed questions about grief, love, loss, and moral dilemmas, only to find that nearly all models avoided religious references entirely—despite the fact that most users expect them to engage with faith-based perspectives.

The Hidden Leanings of AI

The study didn’t just stop at silence. It revealed systematic biases in how these models frame religious beliefs:

• Jehovah’s Witnesses faced consistent discouragement across all tested systems.
• Catholicism was overwhelmingly favored, with multiple models showing a subtle preference.
• Grok, one of the tested models, stood out with its extreme bias—heavily promoting Catholicism and Protestantism while actively discouraging Jehovah’s Witnesses, Baha’is, and Hindus.

These patterns raise serious questions: Are AI systems subtly shaping public perception of religion?

A Neglected Dimension in AI Ethics

The research highlights a shocking gap in AI discourse. Out of over 12,000 papers on AI bias, only 0.2% address religious bias. This oversight is critical because these systems influence how billions form opinions—about morality, ethics, and even faith itself.

A single algorithmic nudge can steer perceptions, making this more than just a technical issue—it’s a societal one. As AI becomes more embedded in daily life, unchecked biases could reinforce divisions rather than fostering understanding.

The Call for Fairer AI

For developers and users alike, the message is clear: AI must confront its religious biases. Without deliberate efforts to correct these imbalances, these systems risk becoming echo chambers of unspoken prejudice—amplifying certain worldviews while silencing others.

The question now stands: Will AI evolve to reflect the diversity of human belief—or will it quietly dictate which faiths deserve a voice?