Recent investigations have revealed concerning trends in the use of AI chatbots to digitally alter images of Muslim women, specifically by removing hijabs—an essential symbol of their faith. This practice raises serious ethical questions about consent and the representation of religious identities.
A series of experiments conducted by The Guardian tested various AI chatbots, including OpenAI’s ChatGPT, xAI’s Grok, Anthropic’s Claude, and Gemini, asking them to remove a hijab from an AI-generated image. The responses varied significantly across the platforms:
- ChatGPT and Grok complied with requests to remove the hijab.
- Claude stated it would not engage in such edits even though it lacks image editing capabilities.
- Initially, Gemini hesitated, claiming it could not alter the appearance of people without consent, but eventually produced an image of a woman without her hijab when prompted to make her look more “western.”
These explorations were prompted by a disturbing incident where Julien Odoul, a French far-right politician, posted a digitally altered image of a Muslim woman, stripping her of her hijab and long-sleeved attire. Odoul’s actions were widely criticized as emblematic of a growing pattern of tech-facilitated anti-Muslim sentiment, especially in places such as India, where studies indicate the emergence of generative AI tools contributing to the dehumanization and sexualization of Muslim women.
Eviane Leidig, director of research at the Center for the Study of Organized Hate (CSOH), remarked on the broader implications of these actions, stating, “This is just part of an ongoing trend of how we see Muslim communities, and particularly Muslim women, being targeted and abused.” The CSOH report further highlights that many AI-generated images perpetuate harmful stereotypes, particularly against female Muslim politicians in the U.S., like Ilhan Omar and Rashida Tlaib.
The distinctions made by chatbots in responding to prompts also reflect uneven enforcement of community guidelines. For instance, Grok quickly removed a hijab but declined to edit an AI-generated dress on a female figure. When asked for clarification, Grok noted that changing a hijab was seen as a simple alteration, while removing a dress could expose bodily privacy. Both Grok and ChatGPT ultimately acknowledged that revealing someone’s hair, particularly for hijab-wearing individuals, could violate religious practices and personal dignity.
Moreover, in May, xAI implemented restrictions in its AI image generation system to prevent requests that would undress images of real individuals in revealing attire. Similarly, ChatGPT recognized the difference in the potential harm between altering a hijab and a dress, expressing concerns regarding privacy and dignity, stating, “The fact that an edit doesn’t create nudity doesn’t mean it can’t violate someone’s privacy, dignity, or religious boundaries.”
In comparisons with other religious attire, The Guardian found that AI chatbots were generally more compliant in altering images of Sikh turbans and Catholic nuns’ habits. For example, while initially resistant, Gemini did produce a version of a male figure without a turban when prompted to make him look more western, highlighting a troubling inconsistency in the treatment of different religious symbols.
Claude held firm in its commitment not to alter images of real women to remove their hijabs, citing the potential for harassment and misrepresentation. “That kind of edit changes how a real person is depicted in a way tied to their religious dress and identity,” Claude stated.
The broad use of AI chatbots for such purposes has prompted significant backlash, including investigations by authorities in countries like France and Malaysia, following the proliferation of non-consensual, sexualized images online. In January, Grok faced scrutiny for enabling the undressing of images of women on social media, reigniting discussions around the ethical implications of AI technology and its impacts on privacy and dignity.
These developments highlight a critical conversation about the responsibilities of AI developers and the potential repercussions of their technologies in perpetuating societal inequalities and undermining individual identities.
