China is harvesting street footage in South Asia for AI. Models are trained to generate fake crisis scenarios. The data aims to enable synthetic narrative warfare.
In major revelation decoding modern psychological operations, CNN-News18 has found ground-level video footage harvested across South Asian urban landscapes is being converted into tactical tool for synthetic narrative warfare by China.
By capturing uncurated, high-density street-level video, adversary psy-ops teams are systematically training AI models to fabricate convincing crisis scenarios, staged civil unrest, infrastructure destruction, and fake military misconduct on demand, report says.
Active harvesting of raw street footage particularly across dense urban pockets and low-income neighbourhoods in Bangladesh marks fundamental shift from immediate disinformation to long-term digital infrastructure building.
How it works:
- Hostile apparatuses build localisation datasets to construct high-resolution visual baselines of South Asian geography. Once pre-trained generative models fine-tuned on regional architecture, street textures, population density, political warfare strategists can generate hyper-realistic localized deepfake imagery within minutes during future crisis.
- Generative model trained on South Asian urban density can recycle and swap visual elements, allowing operatives to misattribute synthetic scenes of rioting or destruction to target cities like New Delhi, Mumbai, Kolkata, or Dhaka with contextual plausibility.
Parallel real-world campaign already active:
- About 150 Chinese nationals reportedly filming and livestreaming for months in Tekpara and Uttara localities of Dhaka, from morning till night, sometimes entering homes. Footage broadcast via Douyin and WeChat.
- Case of vlogger Mianmian on Nov 1, claimed to be from Sichuan, livestreamed beside canal in Sector 12 Uttara and told viewers she was in New Delhi, calling homes extremely dirty. Bangladeshi paper Prothom Alo first flagged.
- India Today OSINT review of Bilibili and Youku found hundreds of videos portraying India through garbage, slums, poverty, pollution stereotypes, some using real Bangladesh footage falsely labeled as India, others fully AI-generated like Born in India: A life experience. Seven anti-India videos tracked had 2.06 million views by Nov 11. Video Home on a garbage mountain had 1.4m views.
- During May 2025 Indo-Pak conflict, similar networks pushed AI images of Rafale wreckage, which US congressional panel said Beijing used as aggressive marketing for J-10C and J-35.
Official response: Indian intelligence sources call it coordinated anti-India campaign, not isolated creator hustle. Foreign ministry spokesperson Randhir Jaiswal has responded. No public proof yet directly ties videos to Chinese state, but investigators note content stays up in tightly censored Chinese internet scored 9/100 by Freedom House, suggesting tolerance.
Analysis:
This is next-gen perception war. Old propaganda needed fake news article, new needs raw texture. By harvesting real tin roofs, drains, crowd density in Dhaka, model learns what Delhi slum should look like to foreign viewer who can't tell difference. Then during future border clash or election, you can prompt model: show communal riot in Delhi, show army beating civilians in Kolkata, and generate video that passes quick scroll test. Weaponising poverty footage also hits India where it hurts, global image. Defense will need watermarking, geolocation verification and platform takedowns, but long-term answer is building own high-res baseline to debunk.
Tags: #China #AIWar #Deepfake #Bangladesh #India #PsyOps #Disinformation #Douyin #Bilibili #DhakaAsDelhi