Bengaluru techie creates realistic PAN, Aadhaar card for 'Twitterpreet Singh' using Google's Nano Banana; leaves netizens divided

A Bengaluru-based techie left the internet divided after he used Google’s AI tool Nano Banana to generate a realistic-looking PAN card and Aadhar card.

Harveen Singh Chadda shared the generated images on X (Twitter) and highlighted a potential issue. In the post, he said, "Nano Banana is good, but that is also a problem. It can create fake identity cards with extremely high precision. The legacy image verification systems are doomed to fail."

The images were of a highly realistic-looking PAN card and Aadhaar card, which belonged to “Twitterpreet Singh”.

Chaddha’s post sparked conversation among users about how the advanced generative AI tool could be misused to create fake identity card documents, which could pose a potential threat to security frameworks.

Some users were not convinced that it could pose a threat. One person noted, “But the number will not be matched and a QR-like unique pattern, and if that system is not able to identify it, then what type of garbage system is that.” To that, Chadda replied, “Tell me when you show Aadhar at a hotel or at airport, do they really scan it ??"

Users also said that the images contained SynthID markers, which could easily be detected. Synth ID markers are a watermarking tool used by Google to mark AI-generated content.

One user said that creating fake IDs was not something new and that it was always possible with Photoshop.

Another person said, “Scanning the QR on the Aadhar and actually verifying it will start becoming the norm soon because of this. And the new Aadhaar app also makes it easy for private hotels and places to verify an Aadhaar themselves.”

A closer look at the IDs shows that the they both have a Gemini AI watermark.

Chadda explained that the intention behind his post was to create awareness “A lot of people are scared but the intention behind the post was not to create fear; it was to create awareness. I wanted to show what today’s AI models are capable of because they work unbelievably fast and with far fewer errors than older methods. This also calls for improving our traditional verification systems at the same pace these models are improving.”