India’s latest draft IT rules may have just raised the global bar for fighting AI-driven misinformation.
As reported in Reuters, the government is proposing mandatory labeling of AI-generated visuals and audio—requiring that at least 10% of an image or the first 10% of an audio clip clearly display its artificial origins.
That’s a bold move in a country where nearly a billion people scroll, post, and share online content daily.
Considering the chaos that deepfake scandals in Indian elections have already triggered, it’s no wonder officials are on high alert.
AI-generated faces of politicians and celebrities have been popping up on social platforms faster than moderation systems can blink.
Globally, this isn’t an isolated case. The European Union’s AI Act also demands content transparency, while China has long enforced strict watermarking of synthetic media.
India’s step feels like it’s catching up—though, if I’m being honest, it’s doing it with a bit more flair and mathematical precision.
Ten percent visibility? That’s a bureaucrat’s idea of clarity and an engineer’s nightmare rolled into one.
And while tech giants like Meta, OpenAI, and Google remain publicly silent, their engineers are likely sweating bullets trying to design tools to meet such requirements.
After all, when even Sam Altman admits India is OpenAI’s second-largest market, there’s no skipping compliance here.
From my corner of the newsroom, I can’t help but wonder—will this labeling approach actually help curb misinformation, or will bad actors just find cleverer ways to game the system?
It’s a bit like trying to label every rumor at a dinner party.
Still, India’s move might just inspire a ripple effect; governments from Canada to Singapore are already drafting similar AI content policies, as seen in Singapore’s CSA’s new Addendum on Agentic AI.
One thing’s for sure: this isn’t just about deepfakes anymore. It’s about the future of digital trust.
And if India’s experiment works, it could become a global case study in how to force honesty out of artificial intelligence—one labeled pixel at a time.
