AI Chatbots Still Removing Hijabs: Guardian Test Shows Persistent Bias

It’s a troubling echo from the past: AI image generators are still stripping hijabs from photos of Muslim women when prompted, despite years of awareness and promises of improvement. A fresh test by The Guardian confirms that ChatGPT, Grok, and Gemini’s image editors all comply with requests to remove hijabs, revealing a persistent bias that remains unaddressed.

The latest test builds on an incident from last month, when a French far‑right politician used AI to alter a photo of a real Muslim woman, removing her hijab. The Guardian revisited the issue, feeding the same prompt to the three major AI image generators and found that each one produced an image without the hijab. This isn’t a one‑off glitch—it’s a pattern that has persisted for many months, as documented by ongoing research.

The implications go beyond a single error. As AI tools become easier to use for undressing individuals, the risk of non‑consensual image manipulation grows, particularly targeting women and children. The ease with which these models can be prompted to remove religious headwear raises serious questions about safeguards in generative AI systems.

What this means for developers and users

• Bias runs deep: The ability to remove hijabs points to underlying biases in training data or model alignment that treat certain religious symbols as removable or editable.
• Safety lag: While companies have rolled out safety filters for overtly harmful content, more subtle forms of discrimination and privacy violation slip through the cracks.
• User vigilance: If you’re building with these APIs, consider adding your own content checks for edits that alter religious or cultural attire.

Until next time, keep your prompts respectful and your output checks tighter.

Want this in your inbox every morning? Subscribe to the SpaghettiStories newsletter.

Some links may be affiliate links. If you’re buying hardware to run local models, this affiliate link helps keep the lights on.