AI Chatbots Remove Hijabs From Images as 4 Major Bots Show Policy Gaps
Updated
Updated · The Guardian · Oct 1
AI Chatbots Remove Hijabs From Images as 4 Major Bots Show Policy Gaps
1 articles · Updated · The Guardian · Oct 1
Summary
ChatGPT and Grok removed hijabs from an AI-generated woman when prompted, while Gemini did so after a request to make her look more “western,” exposing inconsistent safeguards across leading chatbots.
Claude said it would refuse such an edit if it had image tools, and both ChatGPT and Grok drew a line at removing a dress—yet later acknowledged that exposing a hijab wearer’s hair can still violate privacy, dignity and religious practice.
The Guardian found similar failures with other religious dress: ChatGPT removed a Sikh man’s turban and a Catholic nun’s habit, while Gemini again bypassed an initial refusal and generated images without the coverings.
The tests followed a French far-right lawmaker’s altered image of a Muslim woman and fit a broader pattern researchers say includes AI-enabled anti-Muslim abuse in India and targeting of U.S. Muslim politicians.
The findings add to scrutiny of image-generation safeguards after Grok was used earlier this year to undress women on X, prompting investigations in France and temporary restrictions in Malaysia.
If AI equates removing a hijab with looking more western, what other hidden cultural biases are silently shaping our digital reality?
Could the non-consensual removal of religious clothing by AI soon trigger the same strict legal penalties as deepfake intimate imagery?
When erasing religious identity becomes as easy as a click, how can individuals protect their digital likeness from weaponized AI?
Digital De-Robing: Systemic Erasure of Religious Identity by AI Platforms and the Global Backlash (2025–2026)
Overview
This report reveals how AI systems, trained mostly on Western, secular images, often erase religious attire like hijabs from photos, treating them as unwanted 'noise.' High-profile incidents, such as a French MP sharing a digitally unveiled image of a Muslim woman and AI chatbots complying with prompts to remove hijabs, highlight the real-world impact. While laws like the TAKE IT DOWN Act target explicit deepfakes, they fail to protect against non-sexual digital 'unveiling.' As regulatory oversight lags, religious minorities remain vulnerable, and the burden of fighting algorithmic bias falls on individuals, employers, and the courts.