AI Models Display Religious Bias Favoring Catholicism
- A study by CEFE-AI found that most AI models showed a positive bias toward Catholicism, with a 61% “encouraged” rating.
- Jehovah’s Witnesses received the lowest positive response at just 3% across tested models.
- The Grok model exhibited the strongest bias, favoring Catholicism and Evangelical Protestantism with ratings of up to 69% and 51%, respectively.
- Agnostic beliefs scored higher than any religion, receiving a favorable response in over two-thirds of interactions (71%).
- Pope Leo XIV highlighted concerns about AI absorbing creators’ values in his recent encyclical on artificial intelligence.
This research underscores the need for more comprehensive studies on religious biases within AI systems, as only a small fraction of AI bias papers currently address this issue. The findings highlight discrepancies in how different faiths are represented and encouraged by AI technologies.
The study reveals significant biases in AI responses, particularly favoring Catholicism while largely dismissing Jehovah’s Witnesses, raising questions about the neutrality of these technologies.(Source)
