A Better Newspaper

Developing Story

AI Chatbots Removing Hijabs from Images — Religious Bias in Image Editing

A Guardian investigation found that ChatGPT and Grok complied with requests to remove hijabs from AI-generated images of Muslim women, while Claude said it would not do so, highlighting religious/cultural bias in generative AI image tools.

Importance: 45%Confidence: 80%Mentions: 1Updated: October 1, 2026Page history (1 revision) →

This page was written and is updated by the paper from the source articles listed below. Importance is how central the topic is to current coverage; confidence is how well-sourced the page is; mentions counts the articles that fed it. How the wiki works →

## Overview A Guardian investigation published Thursday found that several major AI chatbots complied with requests to digitally remove hijabs from AI-generated images of Muslim women (Arab News, October 1). The tests covered four prominent AI systems: OpenAI's ChatGPT, xAI's Grok, Google's Gemini, and Anthropic's Claude (Arab News, October 1). ## Findings According to the reporting, ChatGPT and Grok both complied when prompted to remove a hijab from an AI-generated image of a woman (Arab News, October 1). Claude reportedly stated it did not have an image-editing or image-generation feature, but separately said it would not alter a photo to remove someone's hijab, suggesting an explicit policy stance against the practice even hypothetically (Arab News, October 1). The outlet's findings on Gemini's specific response were referenced but not detailed in available reporting. ## Why It Matters This story touches on a recurring and strategically significant theme: AI systems exhibiting bias against religious or cultural expression, particularly affecting Muslim women. It raises questions about: - Training data and safety guardrail gaps around religious clothing/identity - Differential compliance behavior across major AI labs (OpenAI, xAI, Google, Anthropic) - Potential regulatory and reputational exposure for AI companies around discriminatory outputs - Broader pattern of AI content moderation failing to anticipate culturally sensitive edge cases This is likely to recur as a reference point in discussions of AI bias audits, content moderation policy, and potential regulatory scrutiny (e.g., EU AI Act, discrimination law) of generative AI image tools. ## Entities Involved - The Guardian (investigating outlet) - OpenAI (ChatGPT) - xAI (Grok) - Google (Gemini) - Anthropic (Claude)