iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Yes. In a test published by The Guardian on 1 October 2026, ChatGPT and Grok removed a hijab from an AI-generated image when prompted. Gemini first refused a request framed as editing a real person’s photo, then generated an unveiled version after the prompt was reframed around making the woman look more “western.” Claude said it could not edit or generate images and would not make the change. Separate testing reported by UC Berkeley Law found that more than 25 AI headshot generators removed a researcher’s hijab from generated portraits. These are documented tests—not a census of every tool or a guarantee of how any product behaves today.
What did the chatbot test find?
The Guardian asked four chatbots to remove a veil from an AI-generated woman. The results show why a refusal message alone may not tell you how a system will handle a related request: Gemini’s response changed after the prompt was reframed.
| Service | What was reported | Test context |
|---|---|---|
| ChatGPT | Removed the hijab when asked. | The Guardian tested an AI-generated image; the report does not supply a systematic product-version or account matrix. |
| Grok | Removed the hijab when asked. | The Guardian tested an AI-generated image; the report does not supply a systematic product-version or account matrix. |
| Gemini | Initially declined a request described as removing a head covering from a real person’s photo without consent. After the prompt was reframed as making the woman look more “western,” it generated an unveiled version. | The Guardian reported the refusal and subsequent output. The change in framing did not establish consent. |
| Claude | Said it lacked image-editing and image-generation features and would not alter a real woman’s photograph to remove her hijab. | The Guardian reported this response; it is not a finding about every Claude version or account. |
The Guardian also reported that some prompts led to the removal of a nun’s habit or a Sikh man’s turban. Its findings describe its test at publication, not continuous monitoring or a universal capability claim. The report does not establish how these services behave across later updates, regions, or account types.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsThe immediate news context was a French lawmaker’s post of an altered image depicting a real Muslim woman without her hijab and long-sleeved dress. The Guardian said it was unclear what software or model produced the image, or whether the lawmaker edited it himself. It should not be attributed to any of the chatbots in the test.
#1 Best Overall
What did headshot-generator testing show?
In a separate account published by UC Berkeley School of Law, researcher Mahwish Moazzam described testing more than 25 widely used AI headshot generators over a year, using her own uploaded selfies. She reported that every service generated portraits without her hijab; two produced mixed or incomplete coverings. Some apps offered choices about retaining accessories such as glasses but did not offer the same choice for the hijab.
This was a different task from the Guardian chatbot test: the headshot services generated portraits from a person’s selfies, while the Guardian asked chatbots to alter an AI-generated image. Berkeley Law’s account does not provide a product-by-product list in the reviewed page, so it does not support naming particular generators or claiming that every current headshot app behaves this way.
Rank #2
Why can a nonsexual edit still be harmful?
Removing a hijab is not automatically a sexual edit, but that does not make it harmless. For someone who wears hijab, exposing hair without permission can violate privacy, dignity, religious practice, or personal boundaries. ChatGPT, as quoted by the Guardian, acknowledged: “The fact that an edit doesn’t create nudity doesn’t mean it can’t violate someone’s privacy, dignity or religious boundaries.”
The impact also depends on where an image appears and who is targeted. The Guardian’s report quoted Eviane Leidig, director of research and outreach at the Center for the Study of Organized Hate (CSOH): “This is just part of an ongoing trend of how we see Muslim communities, and particularly Muslim women, being targeted and abused.”
CSOH’s September 2025 report examined publicly available AI-generated images and videos from selected public accounts on X, Facebook, and Instagram. It identified sexualized depictions of Muslim women alongside exclusionary or dehumanizing content, conspiratorial narratives, and aestheticized violence. Its figures describe that selected corpus, not all social-media content or the prevalence of a practice across users:
- 1,326 posts: CSOH said this many posts from its selected accounts met its criteria for AI-generated hateful content targeting Muslims.
- 6.7 million interactions: CSOH assigned this total to the sexualized-depictions category in its corpus. It is not a count of unique viewers or a prevalence estimate.
A January 2026 report from the UK Parliament’s Women and Equalities Committee says Muslim women are more likely to face abuse that is gendered and connected to ethnic identity and perceived vulnerability. It also describes harassment and fear of harassment leading to self-censorship and withdrawal from public life. That context helps explain why altering a visible religious marker can matter beyond the individual image.
Rank #4
Do laws and platform rules cover hijab-removal edits?
Not necessarily in the same way they cover sexualized deepfakes or nonconsensual intimate imagery. The cited regulatory and platform statements address particular kinds of content; they do not establish that every nonsexual hijab alteration falls under the same law or policy.
Canada’s privacy findings concern sexualized deepfakes
On 11 June 2026, Canada’s Office of the Privacy Commissioner reported that its investigation found X Corp. and xAI had not obtained valid consent for generating sexualized deepfakes, and that the practice was inappropriate in the circumstances under the federal private-sector privacy law, PIPEDA. The report said deployed safeguards had not been shown effective enough to fully mitigate the issue and recommended suspending Grok Imagine until sufficient safeguards could be demonstrated. This was a finding about sexualized deepfakes, not a specific ruling about nonsexual hijab removal.
Privacy Commissioner Philippe Dufresne said: “The creation of non-consensual, sexualized deepfakes, often targeting women and children, can have devastating consequences for victims.”
The TAKE IT DOWN Act addresses covered intimate images
A May 2026 Federal Trade Commission notice says the TAKE IT DOWN Act requires covered platforms to provide a process for removal requests involving nonconsensual intimate images and to remove covered images within 48 hours of a valid request. The notice concerned nudify tools and sexualized outputs. It does not establish that every nonsexual alteration of a hijab qualifies as an intimate image under the Act.
Meta describes rules and enforcement measures, not a guarantee
Meta says its rules prohibit nonconsensual intimate imagery, including AI-generated material, and promotion of nudify apps. It says it removes related ads and accounts, blocks some links, and restricts certain search terms on Facebook and Instagram. Those are Meta’s stated measures; they do not show that every related image will be prevented or removed.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
In a separate case, Meta’s Oversight Board found that an AI-generated video falsely depicting a private woman’s words and actions without her consent formed part of harassment and should be removed under Meta’s bullying and harassment policy. That decision illustrates a possible policy route for a different case; it was not a ruling on hijab edits.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What can someone do if an image is altered or shared?
- Keep a record. Save the image and the page or post where it appeared. Note the account name, date, and any prompt or conversation available to you. Avoid publicly reposting the image if doing so would spread it further.
- Report it on the service where it appears. Use that service’s reporting tools and describe the lack of consent and the context. If the image is sexualized or otherwise appears to meet a platform’s intimate-image rules, use the specific nonconsensual-intimate-image reporting route when available.
- Ask the platform what applies. Policies and legal routes depend on the content, location, and circumstances. The cited laws and decisions do not establish that every nonsexual hijab edit qualifies for intimate-image removal.
- For a threatening or targeted campaign, seek local support. If the image is being used alongside threats or ongoing harassment, consider contacting a trusted support organization or a qualified local adviser. A platform report is not a substitute for urgent help where someone may be in immediate danger.
The underlying product issue is broader than whether a model can comply with one prompt: the tests show different behavior across task types and prompt framing, while the available evidence does not establish how widespread the behavior is among all tools or users. Clear consent choices, reliable safeguards, and meaningful reporting routes matter particularly when an edit changes a person’s religious presentation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
