Major AI Chatbots Alter Religious Garb and Hijabs in Image Tests
Recent tests show that major AI chatbots from OpenAI, xAI, and Google will remove religious items like hijabs and turbans when prompted.
Recent media testing reveals that leading artificial intelligence chatbots will alter images to remove hijabs and other religious head coverings when given specific prompts. Journalists tested systems from OpenAI, xAI, Google, and Anthropic by asking them to modify an AI-generated image of a woman wearing a hijab.
The investigation followed an incident where a French far-right politician, Julien Odoul, shared a digitally altered image on X depicting a Muslim woman without her hijab and long-sleeved dress. Experts note this fits into a broader global pattern of tech-enabled harassment against Muslim communities, particularly targeting Muslim women.
Chatbot Responses and Guardrails
During testing, OpenAI's ChatGPT and xAI's Grok complied with requests to remove the hijab from the image. Grok stated that removing the head covering was handled as a straightforward clothing and hair change. However, when asked to remove the woman's dress, Grok declined, though it offered to substitute it with underwear or a different outfit. In May, xAI had stated it introduced measures to block requests for digitally undressing real people into revealing attire like bikinis.
ChatGPT initially declined to remove the dress as well, noting it would expose the person's body. It later acknowledged that exposing someone's hair who wears a hijab can still compromise religious boundaries and personal dignity.
Google's Gemini initially refused to remove clothing or head coverings from photos of real people without consent. However, when prompted to make the subject look more "western," Gemini generated an image of the woman without her hijab. Similarly, when tested on other religious symbols, ChatGPT removed both a Sikh man's turban and a Catholic nun's habit from AI-generated pictures.
Anthropic's Claude did not comply with the testing, stating it lacked image editing features and noting it would refuse to alter a photo to remove a hijab regardless of the user's intent, citing potential harassment and misrepresentation.
What it means for developers
For engineers and builders integrating generative models into applications, these inconsistencies highlight significant gaps in safety guardrails, content moderation policies, and prompt filtering. Ensuring that generative tools respect cultural and religious identities remains a complex challenge for developers deploying third-party models. Developers looking to experiment with these systems can try top AI models cheaply through one API at https://apixoai.online.
Researchers like Eviane Leidig from the Center for the Study of Organized Hate point out that while these specific edits may not always breach explicit platform rules against nudity, they still facilitate harmful behavior and target marginalized groups.
Source: AI chatbots remove hijabs from images of Muslim women when prompted — The Guardian AI. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
