In this Blog: Let’s discuss about Nano-Banana, its origin and its collision with the DPDPA.
Image Generation via AI started to become an obsession with Chat-GPT introducing the Ghibli Trend. Next obsession came in the form of 3D sculpture trend all over, followed by the Saree trend and now the Retro Baby Selfie …
Where did these trends come from? How did become India’s new obsession and which is this image generation model with a weird name? Was Nano Banana always there disguised under a different name or a latest model by Gemini.
In the past few blogs we have discussed about Large Language Models, which form the basis for generating the output in form of images. Yesterday, Josh Woodward, Vice President at Google announced that “its Gemini 2.5 Flash Image model, known as Nano Banana, has generated over 5 billion images in under a month, becoming a social media trendsetter for retro selfies and 3D models”.
Nano Banana first popped up on a site called LMArena, a place where different AI models compete anonymously in a “Battle Mode.” (Source: Medium.com)
The origin of Nano Banana was initially questioned and everybody had their own theories about who it belonged to a few weeks back; when Google had not specifically announced that it was owned by them. It first surfaced on LMArena, a competing platform, where one prompt is given to two anonymous models and which one generates the best result is looked for. Later, it started showing up on Discord and other AI testing platforms. It was Google’s little secret. Google was trying out its big step in the generative media.
Distinctiveness: Nano Banana is the image generating and editing model by Google DeepMind. What makes it distinctive from other image generating models is its ability to- make complex edits to images, like removing objects from the background or changing the background altogether or a subject’s clothes, while keeping the overall image coherency intact. This is where we get the Saree trend coming from.
Another point in its distinctiveness is it only needs your prompts and not your skills! Once the prompt has been fed, the rest is taken care of by the model. Secondly, the person or the object that you want to common stays as it is while he background, angles, colors can be changed easily. Lastly, it is super fast and takes around 15 seconds to generate the image and allows multiple usages of the same.
Nano Banana and The DPDPA:
Every AI Image Generation trend, would always collide with the ‘Consent’ under the Digital Data Protection Act. May it be Ghibli or these new ones, the tussle of ‘explicit consent’ remains a problem.
Users who are commoners, and not related to the fields of AI, often fail to understand the risks that are posed by the usage of such models. Often in my earlier blogposts I have mentioned how Generative AI (LLMs) collect the datasets that are publicly and the prompts available and generates the desired output.
Consent under DPDPA has to be FREE, SPECIFIC, INFORMED, UNCONDITIONAL, UNAMBIGUOUS and REVOCABLE. None of these is present while a user lacking the basic understanding of privacy risks, provide a photo of themselves to the models. The grey area of accountability of AI and classification of AI as either being an intermediary or a platform, makes it more difficult to put a statutory power on it.
Such image generations are a critical violation of the DPDPA and also unintentional privacy violation of the users done by their own selves. Even a lot of tech-savy people turn a blind eye while providing loads of personal data to the LLMs. Till the time, we do not have a proper intersection of legal regime with the AI space, using Generative AI carefully and with an open eye is a must!
YLTB! 
Leave a comment