India's response has been proactive but incomplete. The rapid succession of IT Rules amendments in 2025 and 2026, mandating labelling, metadata embedding, and 36-hour (and now 3-hour) takedown windows, represents one of the most aggressive regulatory stances on deepfakes anywhere in the world. Yet experts caution that such tight deadlines may lead to automated over-removals without adequate human review.
While deepfakes are a global issue, the impact of desifakes is amplified by specific cultural dynamics within South Asian societies:
Creating comedic parody videos of politicians interacting in fictional, lighthearted scenarios. 2. Political Misinformation and Propaganda desifakes ai generated
A particularly egregious case involved a video clip of Shah Rukh Khan promoting a fake 'passive income' app allegedly linked to Mukesh Ambani, which was framed as a news segment featuring a real journalist and a political leader. Fact-checkers later confirmed the voices had been fabricated and the website linked was fraudulent. By the time the video reached a newsroom, it had already flooded WhatsApp groups across the country.
Over the last decade, millions of users across India, Pakistan, Bangladesh, and the global diaspora gained access to high-speed internet and cheap smartphones. India's response has been proactive but incomplete
Tech companies and researchers are developing AI-based deepfake detectors. These tools analyze digital media for anomalies that the human eye might miss, such as unnatural blinking patterns, irregular lighting on the skin, or audio-visual desynchronization. Platforms like Google, Meta, and X (formerly Twitter) are increasingly implementing automated systems to flag, label, or remove synthetic media that violates their terms of service. 2. Legal Frameworks and Regulation
Swapping Hollywood actors into iconic Bollywood movie scenes. Reviving deceased musicians to "sing" modern hit songs. While deepfakes are a global issue, the impact
To understand the threat, one must understand the accessibility of the tools. Five years ago, creating a convincing face-swap required a powerful GPU, thousands of images, and expertise in machine learning frameworks like TensorFlow or DeepFaceLab.
Generic deepfake models are trained on Western datasets. However, "DesiFakes" vendors have fine-tuned their models to understand South Asian nuances:
