Artificial intelligence tools are being used to fabricate and circulate sexualized imagery of Muslim women in India, a targeted form of online harassment that exploits the accessibility of generative AI to inflict reputational and psychological harm on a religiously marginalized community, Al Jazeera reports.

The mechanics are not novel in a global context — non-consensual deepfake pornography has been documented across multiple countries — but the India case carries a distinct communal dimension. The targeting is not incidental or random. Muslim women are being selected because of their religious identity, placing this pattern within a broader spectrum of digitally enabled communal violence that has intensified alongside rising Hindu nationalist sentiment. The images are designed to humiliate, silence, and intimidate. That the targets are women compounds the harm: gender and religious minority status interact to create a specific vulnerability that general anti-harassment frameworks rarely address with precision.

The Technology and Its Misuse

Generative AI image models — trained on vast datasets and now accessible via consumer applications requiring minimal technical skill — can produce photorealistic imagery indistinguishable from authentic photographs. That accessibility is the operative factor here. Earlier iterations of image-based abuse required either stolen explicit material or rudimentary face-swap software that produced visibly artificial results. Current tools close both gaps: no source material is needed beyond a publicly available photograph, and the output quality is high enough that targets describe it as looking "so real."

The harassment infrastructure layers on top of pre-existing social media amplification dynamics. Fabricated images can be seeded into WhatsApp groups, Telegram channels, or public social media accounts and spread faster than any takedown mechanism can respond. India's internet user base — now among the world's largest — means the potential for rapid, wide distribution is structurally embedded in the problem.

Legal and Institutional Gaps

India's legal framework is ill-equipped for this specific threat vector. The Information Technology Act and its amendments address certain categories of online obscenity and non-consensual image sharing, but prosecution under these provisions has been inconsistent and slow. Identifying perpetrators operating through anonymized accounts across jurisdictions is a resource-intensive process that Indian law enforcement has historically deprioritized in cases involving minority complainants.

Platform liability is the other pressure point. The major AI image-generation services operate predominantly outside Indian jurisdiction, complicating both takedown requests and evidentiary access. India's IT Rules of 2021 impose due-diligence obligations on significant social media intermediaries, but enforcement against foreign-headquartered AI tool providers remains largely theoretical.

The absence of a dedicated deepfake-specific legal provision is not unique to India — most democratic jurisdictions are still drafting or debating such laws — but the communal targeting dimension makes the gap more acute. Hate speech and targeted harassment of religious minorities theoretically fall under existing Indian Penal Code provisions, yet successful prosecutions in digitally mediated communal harassment cases are rare.

Why This Matters Beyond India

The pattern is a data point in a broader global trend: AI-enabled image abuse being applied not just to celebrities or ex-partners but to ethnoreligious minorities as a tool of collective intimidation. When the goal is to suppress the public presence of an entire demographic rather than to harm a specific individual, the harm scales differently. Individual takedowns and individual prosecutions address individual instances; they do not address the chilling effect on Muslim women's participation in public and professional life.

That chilling effect is the intended outcome. Silencing by humiliation is an old instrument; generative AI has sharply lowered the cost and raised the reach of deploying it.

The regulatory response globally is still catching up. The EU's AI Act classifies certain deepfake applications as high-risk and imposes transparency requirements, but enforcement architecture is nascent. In India, the gap between the pace of AI tool proliferation and the pace of legislative response is widening. Civil society organizations documenting these cases are, for now, the most active line of accountability — a fragile substitute for structural remedies that have not yet materialized.