I built an AI-driven identity verification startup. But today, I’m deeply worried.
The image below was generated by OpenAI’s latest model, GPT-4o. Yes, that’s a synthetic Aadhaar card — created in under 2 minutes, without writing a single line of code or using any sophisticated tool.
My startup’s mission was to fight fraud and identity theft using AI and we processed millions of IDs. Ironically, we’re now entering an era where AI can also empower bad actors to create hyper-realistic fake IDs at scale. We did build very smart protocols and models to detect fraud but now, the ability to generate synthetic identities is evolving faster than the ability to detect them.
This raises a critical alarm: image-based KYC will soon need a big revamp.
The only viable path forward?
1. Verification through official, cryptographically signed sources like DigiLocker (only).
2. Or the government making Aadhaar verification APIs more accessible, so businesses can triangulate data directly from the source — not just images.
The tools that once helped us fight fraud can now be used to create it.
We’ll need smarter regulation, better public APIs, and stronger protocols to stay ahead.
This post was originally shared by on Linkedin.