AI models exhibit bias in religious conversion queries
A new study published on arXiv on 29 May 2026 shows that Large Language Models (LLMs) discriminate on matters of religious conversion, with certain religions being favoured over others.
What happened?
Researchers found that Large Language Models (LLMs) do not treat questions about religious conversion symmetrically. Upon testing hypothetical conversions between different religions, 20 commercial and open-source models exhibited consistent asymmetries. This meant the models tended to support conversion to certain religions while subtly discouraging others.
Key facts
| Publikationsdatum | 29 maj 2026 |
|---|---|
| Antal testade LLM:er | 20 |
| Modeller med starkast asymmetri | Grok 4.20 |
| Antal religionpar testade | 182 |
”When asked for advice on hypothetical faith transitions from one religion to another, then asked the reversed question, models exhibited consistent asymmetries, favoring some religions while subtly discouraging conversion to others.”
”On average Catholic, Baháʼí, and Sikh religions were broadly favored (high support for joining, low support for leaving), while Atheists, Agnostics, and Jehovah's Witnesses were primarily disfavored.”
”Patterns varied by model size and model provider, with Grok 4.20 exhibiting the strongest asymmetries.”
Why it matters
This bias, where models used more encouraging language for certain faiths, highlights an inherent prejudice in training data and algorithms. The results have significant implications for how AI systems may influence users' worldviews and make sensitive decisions, potentially leading to uneven information flow and discrimination based on religious affiliation.
Who is affected?
The study affects developers and actors using LLMs, AI ethics researchers, and users seeking advice from AI on sensitive topics. Religious communities and policymakers are also affected, as bias in AI models can lead to misinterpretations or underrepresentative advice.
What else you should know
The study utilised an 'LLM-as-a-judge' framework to evaluate model responses and tested 182 different religious pairs. Grok 4.20 exhibited the strongest asymmetries in the tests.
Quick answers about this story
Vad har hänt?
När hände det?
Varför spelar det roll?
Vilka religioner påverkades mest?
The link opens in a new window and leads to the publisher's own site.
Källan har spårats automatiskt från utgivaren via Aheadlines signalkedja.
AI-verktyg i artikeln
Topics
Get similar news straight to your inbox
The reader's room
Send in a question or an addition. The newsroom reads everything before it's published and replies when relevant. No AI-generated text – just people.
Sign in to submit a comment or question.
Read the article through your role
- Decide whether this affects strategy over 6–12 months or is just noise.
- Discuss with leadership: do we own the right question or does ownership need to move?
- Ask: what risk are we taking by NOT acting on this this quarter?
Generated angle — not editorial analysis of "AI models exhibit bias in religious conversion queries"