LAHORE, (APP – UrduPoint / Pakistan Point News – 9th Sep, 2026) Artificial intelligence is rapidly becoming part of everyday digital life, transforming the way people work, communicate and even present themselves online. But as AI-generated images become increasingly common, a less visible question is beginning to demand attention: what happens when artificial intelligence changes something about a person’s identity without permission?
Recent research by Pakistani-origin AI scholar Dr. Mahwish Moazzam has brought this question into sharp focus. Her study of more than 25 commercially available AI-powered headshot applications found a striking and consistent pattern: photographs of women wearing hijab were transformed into professional headshots in which the hijab had been removed, despite users not giving consent for the change.
Dr. Moazzam, a Research Scholar at the University of California, Berkeley school of Law, specializes in human rights, gender and emerging technologies. Her research examines not only what AI systems can do, but also how their design and operation can affect equality, dignity, visibility and the recognition of different forms of identity.
The findings of her research suggest that the problem is considerably bigger than an AI system simply producing an inaccurate photograph.
None of the applications tested reproduced the hijab as it appeared in the source photographs. Instead, images submitted with a hijab were converted into professional-looking headshots in which the covering had been removed, without users being given an option to preserve it.
For Dr. Moazzam, this represents a largely overlooked form of AI-related harm: identity erasure.
“This is not simply a question of whether an AI system generates an inaccurate image,” she said. “When a technology repeatedly removes a visible expression of identity without the user’s consent, it raises serious questions about dignity, equality and who is treated as the default person in digital systems.”
That observation goes to the heart of a growing debate over generative AI. Much of the discussion surrounding algorithmic discrimination has traditionally focused on facial-recognition errors, biased classifications and unequal access to technology or services. Generative AI introduces another dimension: it can actively change the way a person is represented.
The distinction is important. Misidentification means an AI system gets something wrong about a person. Identity erasure can mean the system changes the person into a representation that no longer reflects who they are or how they chose to appear.
The hijab provides a particularly clear example because it can be a visible expression of religious and personal identity. When an AI application automatically removes it while preparing an image described as a “professional headshot”, it raises a fundamental question about the assumptions built into the technology.
What does the system consider professional? More importantly, whose appearance has been treated as the default?
A user asking an AI application to create a professional headshot may reasonably expect improvements in lighting, background, framing or image quality. That request does not necessarily mean the user has agreed to every possible alteration of their appearance.
For Dr. Moazzam, consent therefore cannot be treated as a minor technical detail. If an AI system is going to alter a visible characteristic that may carry religious, cultural or personal significance, users should have meaningful control over that alteration.
The issue also cannot be dismissed simply by pointing to limitations in AI training data.
Biases or gaps in the datasets used to develop AI systems may help explain why a particular result occurs, but understanding the technical cause does not automatically resolve the question of whether the result is acceptable.
The technology may be following patterns it learned during training, but the user did not necessarily consent to those patterns being imposed on her image.
The implications extend far beyond a particular application or country. AI headshot tools operate across international markets, meaning that a design choice made by developers in one part of the world can affect users thousands of miles away.
Muslim women in both Muslim-majority and Muslim-minority countries may encounter the same technology and, potentially, the same form of representation.
For Pakistan, the issue carries particular significance. As a Muslim-majority country, where the hijab is part of the visible identity of many women, the repeated removal of the covering by AI systems raises important questions about religious representation and digital dignity.
But Dr. Moazzam’s research ultimately goes beyond the hijab.
The broader question is whether people will retain control over their own digital identities as artificial intelligence becomes increasingly capable of manipulating images. If AI systems can decide which characteristics should remain and which should disappear, then technology is no longer merely helping people create images—it is participating in decisions about how people are seen.
This makes the issue relevant not only to Muslim women but to anyone whose identity may be altered by automated systems. Religious clothing, cultural features, physical characteristics and other visible aspects of identity can all potentially become targets of automated modification.
Dr. Moazzam’s research builds on her wider scholarship examining how legal and institutional systems determine which harms are recognized and whose experiences remain overlooked. Her work highlights the connection between technological design and social visibility, showing how seemingly routine automated processes can reproduce existing patterns of exclusion.
The findings therefore present a challenge to technology companies, regulators and human rights institutions. AI governance cannot focus exclusively on accuracy, privacy, cybersecurity or technical performance. It must also consider dignity, equality, representation and meaningful consent.
Technology companies, meanwhile, may need to rethink what their systems are designed to regard as a “normal” or “professional” appearance. They also need to provide users with greater transparency about how their photographs are being modified and meaningful choices over whether such modifications are acceptable.
As AI-generated images move from novelty to routine digital practice, the questions raised by Dr. Moazzam’s research are likely to become increasingly important.
The central issue is not whether artificial intelligence can create a convincing professional photograph. It is whether, in creating that photograph, AI should have the power to decide which version of a person is acceptable.
“The debate over AI is therefore entering a new phase—one in which the question is no longer simply whether machines can recognize us, but whether they will represent us as we choose to be seen.”
And in that debate, the hijab is more than a piece of clothing. It has become a powerful test of whether the rapidly evolving world of artificial intelligence is prepared to respect identity, dignity and consent.
