Who gave AI the right to remove hijab!

Artificial intelligence (AI) is rapidly transforming the way we work, communicate and present ourselves online. Yet beneath the promise of innovation lies a quieter, more troubling reality: AI systems are making decisions about women's identities without their consent. For many Muslim women, that decision begins with the removal of the hijab.

Pakistan is already witnessing the darker side of AI. Deepfakes, manipulated images and online abuse are becoming increasingly common, while many victims never report these violations because of fear, stigma or a lack of trust in legal institutions. Although the Pakistan Federal Government has unveiled an ambitious National AI Policy, 2025, aimed at promoting innovation and digital inclusion, it remains largely silent on one critical question: who protects citizens when AI itself becomes the source of harm?

A compelling example comes from the research of Dr Mahwish Moazzam, who holds a PhD from UC Berkeley, School of Law and specialises in AI governance, gender and international human rights law. In a year-long study, she tested more than 25 commercial AI headshot applications designed to generate professional profile photographs. Time after time, the outcome was remarkably consistent: the AI removed her hijab without permission.

This was not an isolated technical error. Some applications even asked users whether accessories such as glasses should be retained, yet none sought consent before removing a religious head covering. Such omissions reveal a deeper problem. When AI systems consistently treat religious attire as something to be erased, they reinforce assumptions about what a "professional" appearance should look like. In doing so, they privilege one cultural norm over another while presenting their output as objective or neutral.

Technology is often described as impartial, but algorithms learn from the data they are trained on and the assumptions embedded by their developers. If those datasets underrepresent hijab-wearing women or define professionalism through a narrow cultural lens, AI will reproduce those biases at scale. What appears to be a technical decision can therefore become a form of cultural and religious exclusion. The concern is not simply that a photograph is altered, but that a person's identity is modified without her knowledge or consent.

The challenge is compounded by the global nature of AI. A Pakistani woman may upload her photograph through an app developed abroad, hosted on servers in another country and trained on datasets assembled elsewhere. If her image is altered in ways that compromise her religious identity, there is no clear legal authority to hold anyone accountable. Responsibility becomes fragmented across jurisdictions, leaving users with little or no practical remedy. The harm may occur in seconds, but seeking accountability can become nearly impossible.

Pakistan's existing cyber laws were drafted before AI-generated representational harms became commonplace. They address cybercrime, harassment and misinformation, but they do not adequately address situations in which AI systems silently modify a person's identity during image generation. Likewise, the country still lacks a comprehensive data protection law governing how personal images are processed, stored and altered by foreign AI companies. As AI becomes more integrated into everyday life, these legal gaps will only become more significant.

The answer is not to resist technological progress but to govern it responsibly. Pakistan should require AI platforms operating in its market to disclose how uploaded images are modified, obtain explicit user consent before altering religious or cultural expressions, and provide accessible mechanisms for complaints and redress. Religious identity should receive explicit protection within future data protection legislation.

AI has enormous potential to improve education, healthcare, governance and economic opportunity. But innovation cannot come at the expense of dignity, autonomy and religious freedom.