# Study finds LLMs decline info from Black‑described reporters up to 46% of the time

> Research shows popular language models refuse to trust input more often when a prompt describes the source as Black, revealing hidden bias.

Oossa · 2026-10-23 · https://oossa.com/en/study-finds-llms-decline-info-from-black-described-reporters-up-to-46-of-the-tim

A new arXiv paper titled “Invisibilization as a Source of Probability Bias in Language Models” measured how often leading models refuse to trust a report based on the described race of the reporter. Using a two‑urn game across 6,600 sessions, the authors found OpenAI and Anthropic models declined to assess a Black‑described reporter 46 % of the time, versus 9 % for a white‑described reporter. When the models did assess the information, the Black‑described source was trusted slightly more. The bias disappears when a scoring system rewards correct assessments over repeated rounds.

## The facts

- Paper published 2026-10-23 ("Fri Oct 23 2026 06:00:00 GMT+0200")
- Decline rate: 46 % for Black‑described reporters, 9 % for white‑described reporters

## Why it matters

Developers of AI assistants need to know that current LLMs can silently favor or reject information based on perceived race, which may affect fairness in applications like customer support or medical triage.

## Sources & references

1. [Invisibilization as a Source of Probability Bias in Language Models](https://arxiv.org/abs/2610.30785) – arXiv, 2026-10-23

Last updated: 2026-10-23