Gemini's 'Nano Banana' turns selfies into AI saree edits; but is it safe?

Nano Banana, Google's Gemini Nano-powered AI photo tool, has gone viral with saree edits but once again sparks privacy and safety concerns over AI image generation

Don't want to miss the best from Business Standard?

This story has been updated.  The new AI-powered photo-editing feature called 'Nano Banana', built on Google’s Gemini Nano model, has taken social media by storm. The tool transforms regular selfies into 3D figurine-style portraits with glossy plastic-like skin, oversized expressive eyes, and playful cartoonish proportions.

 

While users first embraced the quirky figurine look, the trend quickly evolved into the now-viral vintage saree AI edit. The filter reimagines portraits, mostly of women, into glamorous retro-style saree looks, complete with cinematic backdrops reminiscent of classic Bollywood posters. Instagram is flooded with chiffon sarees, flowing drapes, and golden-hour lighting, as millions try the trend.

 

But as with most AI-driven fads, questions about privacy, consent, and data safety are growing.

 

Nano Banana, also known as Gemini 2.5 Flash Image, is the latest AI-powered editing tool from Google. By mid-September, users had already created or edited more than 500 million images in the Gemini app, with hundreds of millions more generated across other platforms, Medium reported.

 

Using the tool is simple: Upload a photo, add a prompt, and watch your image get transformed. 

The creepy side of cute

An Instagram user named Jhalakbhawani recently shared an unsettling experience with the saree trend. She explained that after uploading her photo to Gemini, the generated image showed a mole on her left hand. This is a real detail about her body that did not exist in the original uploaded picture.

 

“How did Gemini know that I have a mole in this part of my body? It’s very scary and creepy,” she wrote. Her post sparked debate in the comment section, with some users speaking about safety concerns while others dismissed it as coincidence or called the post attention-seeking.       Google, however, maintains that any identifying markers in images generated by Gemini are "coincidental".   “The Nano Banana image model that was recently launched was not trained with user data from Google Photos, Google Workspace, or Google Cloud services. Any non-visible attribute that shows up in a generated image that wasn’t visible in an input image in the same Gemini conversation is a coincidence,” a Google spokesperson said.    

 

How safe is Nano Banana?

Google says all images generated or edited using Gemini carry SynthID, an invisible digital watermark, along with metadata tags, to clearly mark them as AI-generated. According to Google’s AI Studio, these identifiers are meant to give creators and platforms a way to verify the origin of content.

 

Google, OpenAI, and xAI (Elon Musk’s AI venture) have also maintained that uploaded images are not permanently stored. However, privacy advocates still insist that users must still tread carefully.

 

Is watermarking enough to safeguard against AI abuse?

However, since the rise of AI images, experts, netizens, and industry experts have been advising caution on this matter. The detection tools needed to read SynthID are not yet available to the public, meaning most everyday users cannot confirm authenticity. Many have also pointed out that watermarks can be easily faked, ignored, or removed.

 

“Nobody thinks watermarking alone will be sufficient,” said Hany Farid, professor at UC Berkeley’s School of Information, in a Wired report. He and others argue that watermarking must be combined with additional safeguards to stand a real chance against deepfakes. In the same article, University of Maryland computer science professor Soheil Feizi said, “We don’t have any reliable watermarking at this point.” 

 

How to use AI image tools safely?

Some safety recommendations made by experts since the mass use of AI image generation include:

 

- Avoid uploading sensitive photos such as intimate images or those showing identifiable personal details.

 

- Strip metadata like location tags before sharing.

 

- Check app permissions and revoke unnecessary camera or gallery access.

 

- Limit exposure by posting low-resolution images instead of high-quality originals.

 

- Read privacy policies carefully to understand how data may be reused or stored.

 

- Specialised tools like Glaze and Nightshade can also add subtle “noise” to images, making them harder to scrape for AI training.

 

Bottom line

While Nano Banana has made AI portrait editing fun, it's important to stay vigilant. Invisible watermarks such as SynthID may offer a starting point for accountability, but they are not foolproof.