Clashes broke out at Ujjain's Shahi Masjid after AI-generated videos showing the entire mosque being demolished spread rapidly across social media. Madhya Pradesh Police later confirmed the videos were fabricated, calling them 'completely false'.
Only a three-foot overhang of the mosque, at the alignment of a road-widening project, was ever sanctioned for removal by the Madhya Pradesh High Court, and even that was being carried out voluntarily by members of the mosque committee themselves. The video in circulation showed something categorically different - total demolition by heavy machinery. That single fabricated visual did more to mobilise a crowd than any actual administrative order did, a reminder that a convincing fake now travels faster and hits harder than the truth it distorts.
Ujjain is not an isolated incident. It is the latest data point in a pattern that has been building in India recently, one where synthetic AI-generated media doesn't just deceive individuals, it moves crowds.
A year of manufactured trust
AI fakes weaponising public trust have hit India's political, financial, and civic life repeatedly in recent months. Earlier this year, a deepfake video of Union Finance Minister Nirmala Sitharaman convinced a Hyderabad woman to invest, and eventually lose Rs. 71.84 lakh in a fake trading scheme. A near-identical scam using her likeness had already defrauded a 76-year-old man in Belagavi of Rs. 7.9 lakh just weeks earlier. In January 2026, philanthropist Sudha Murty had to publicly warn the country that fraudsters were using AI to clone her voice and face to peddle investment scams she had never endorsed.
The fraud has grown more technically alarming too. In April 2026, an Ahmedabad gang was found using Google's Gemini AI to generate deepfakes capable of bypassing Aadhaar-linked biometric verification altogether, defeating OTP checks to push through fraudulent loans. Around the same time, forensic experts publicly called for a dedicated 'AI-driven playbook' simply to keep pace with the scale of deepfake-enabled financial crime hitting Indian investors. A McAfee report from mid-2026 found that roughly 90 percent of Indians had already been exposed to fake celebrity endorsements, with victims losing an average of Rs. 34,500 each, Shah Rukh Khan and Alia Bhatt topped the list of most-exploited faces.
From fraud to physical violence
What sets Ujjain apart from the financial scams is the leap from digital deception to physical mobilisation. A fabricated video didn't just cost people money, it moved a crowd to a religious site and triggered a law-and-order breakdown. That is the genuinely new and dangerous frontier - synthetic media no longer needs to be believed by everyone, it only needs to reach the right pocket of already-anxious people at the right moment. In a country where communal fault lines are historically sensitive and virality outpaces verification, that is an alarmingly low bar to clear.
The common thread across Ujjain, the Aadhaar bypass fraud, and the celebrity investment scams is the same: India's information ecosystem, from WhatsApp forwards to Instagram reels, has almost no friction between a convincing fake and mass belief. Police are left doing cleanup after the damage, arrests, account takedowns, public advisories, rather than intercepting the content before it spreads.
A global pattern, not a local glitch
This is not solely an Indian problem, and it isn't only about doctored images. In Gaza, investigations into Israel's military use of an AI system called Lavender revealed that the tool had, at one stage, flagged around 37,000 Palestinians as potential targets, with intelligence sources indicating a roughly 10 percent error rate, and reviewing officers reportedly spending mere seconds validating each AI-generated recommendation. That reporting continues to shape ongoing scrutiny of algorithmic warfare well into 2026, offering a far more lethal illustration of the same underlying failure seen in Ujjain: humans deferring judgment to a system faster than they can verify it, with catastrophic real-world consequences.
The climate cost
There's a quieter toll attached to all this generative capability, too. AI's environmental footprint is no longer theoretical. Global data centre emissions hit an estimated 286 to 315 million tonnes of CO2 in 2025, according to Allianz Trade, up to 57 percent higher than earlier IEA estimates, with AI workloads alone responsible for a growing share of that number. The same report noted that in India specifically, emissions per unit of electricity used to power this infrastructure run over 20 times higher than in countries like Norway or Sweden, given India's continued reliance on coal-heavy grids. Every deepfake, scam video, and viral clip fuelling India's misinformation crisis is quite literally running on infrastructure with a rising climate bill attached.
What these cases collectively expose is not that AI is inherently malicious, but that verification has not kept pace with generation. A video can be fabricated in minutes, but confirming it is fake takes hours, sometimes days, by which point a mob has already gathered or a woman's life has already been upended. Until platforms, law enforcement, and regulation catch up to the speed at which convincing fakes can now be produced, incidents like Ujjain will not be an aberration. They will be a preview of the norm.
