AI of the Storm: Deepfakes Cloud Protest Coverage

A video of a scantily clad, artificial intelligence (AI)-generated woman trying to stop a police van, sexualised content edited into protest footage, forged documents attributed to protest leaders and old visuals passed off as current footage are among the misleading posts circulating on social media amid the ongoing protests in Delhi-National Capital Region and Mumbai against the chief election commissioner and the special intensive revision of electoral rolls.

Several AI-generated clips have gone viral on Instagram, Facebook and X, drawing millions of views and comments. A recent viral clip of a woman criticising the protests, for instance, led users to an Instagram profile that regularly posts AI-generated political content.

Ankush Tiwari, founder of pi-labs, which builds AI tools for deepfake detection and voice forensics, said that AI tools are making synthetic content faster and cheaper to produce, which makes it difficult for detection systems to keep up. While detection tools can technically scan every video uploaded online, doing so at that scale is impractical, according to him.

"We can't scan everything because in a five-minute call, more than a million content pieces have been uploaded across the globe," Tiwari said. pi-labs looks for pixel-level inconsistencies in AI-generated videos, images and audio to identify potential manipulation, including shifts in audio frequency patterns that can differ from those in natural human speech.

Today's detection systems operate at 80-90% accuracy, and deepfakes have grown to the point where pi-labs is seeing demand for verification surge over 100% every quarter, according to Tiwari.

Experts said the flood of such material has exposed gaps in

content moderation and made it more difficult to tell authentic footage apart from synthetic content. These worries come even as platforms are accused of temporarily blocking the accounts of independent journalists and individuals covering the protests.

Digital rights advocacy group Internet Freedom Foundation (IFF) has sent four notices to Meta seeking the restoration of temporarily suspended accounts belonging to independent journalists. Some of these accounts have since been restored.

Apar Gupta, a lawyer and founder director of IFF, said that the notices sent to affected users cited an automated system implemented by Instagram under the Information Technology Rules, 2021. "The restrictions appear to be a form of proactive moderation by Meta rather than action directly taken under formal government blocking orders," he told ET. "We are also seeing the spread of deepfake content and misinformation about the protests on social media. An information vacuum caused by internet shutdowns and network jamming is making it harder to verify events on the ground."

Meta and X did not respond to ET's queries till press time.

A study published this year and presented at the Computer Vision and Pattern Recognition Conference 2026 found that AI-generated content was disproportionately viral and took longer to be flagged. AI-generated videos took an average of 11.3 hours to receive their first community note, against 9.1 hours for miscaptioned videos, according to the study, titled 'The Synthetic Media Shift: Tracking the Rise, Virality, and Detectability of AI-Generated Content', which analysed more than 150,000 posts on X.

Specialised detectors caught about 75% of AI-generated images in early 2023 but only about 40% by late 2025, the study found.

For Amitabh Kumar, cofounder of Contrails AI, a Bengaluru-based firm that also works on deepfake detection, the trend reflects a wider problem. "Deepfakes are being used in political discourse and conflicts across countries, including the Russia-Ukraine war, US politics and in India for the last few years," he said.

"Deepfakes have emerged as a tool to grab attention, and people in India increasingly use them due to easy access," Kumar added. "Content moderation in India has been appalling since the pandemic and hasn't really improved since then. Social media platforms have also failed to adequately address the problem."

As per McAfee's 2026 State of the Scamiverse survey from February, Indians reported seeing an average of four deepfakes daily, most commonly on Instagram (65%), followed by Facebook (59%), YouTube (48%), Telegram (44%) and WhatsApp (41%). It also found that more than one in three Indians were not confident they could spot a deepfake scam.

Flood of AI material has exposed gaps in content moderation making it difficult to tell authentic footage from synthetic content

(This story has not been edited by economictimes.com and is auto-generated from a syndicated feed we subscribe to.)