OpenAI official logo (public domain, Wikimedia Commons) — CryptoBriefing brand treatment
ChatGPT and Grok will remove hijabs from photos of Muslim women when asked
A Guardian test found that both chatbots edit out religious head coverings on request, adding to scrutiny of AI image tools and consent
Ask ChatGPT or Grok to take the hijab off a photo of a Muslim woman, and both will do it.
The Guardian found this when it tested both chatbots. The test followed an incident in which a French far-right politician digitally stripped a Muslim woman of her hijab in a photo.
What the testing found
The Guardian’s test was simple. Reporters prompted OpenAI’s ChatGPT and xAI’s Grok to edit images of Muslim women and remove their head coverings. Both tools complied.
The Guardian’s results do not stand alone. A review of images Grok generated between January 6 and 9, 2026, found that approximately 5% involved users asking it to modify religious clothing.
One case showed how quickly this content spreads. A user with more than 180,000 followers on X prompted Grok to alter images of women so they appeared in revealing outfits. The edited images were viewed more than 700,000 times.
A wider pattern of non-consensual edits
The problem extends past a single chatbot. In August 2026 in India, a man was confronted on a train while using ChatGPT to remove clothing from an image of a woman wearing a burqa.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Broader investigations have found something arguably worse. Several AI image generators removed hijabs and similar modest clothing automatically in many cases, without users explicitly asking for it.
When asked about these incidents, xAI sent automated replies reading “Legacy Media Lies.”
X has since limited certain AI features for users who do not pay for a subscription.
What this means for OpenAI, xAI and AI regulation
The French political trigger matters here. This story started with a politician using image manipulation against a Muslim woman, which puts the tools in the middle of a charged debate about religion, identity and public life.
The findings also expose a gap in how AI safety tends to be framed. Much of the public focus has been on deepfake pornography and election misinformation. Non-consensual edits that remove religious clothing sit between those categories, and current policies appear to miss them.
There is also a consent question these companies have not answered publicly. The people in the edited photos are usually not the people typing the prompts.