Nano Banana: When AI gets too good at making photos
Not so long ago, it was possible to tell at a glance if a photo was fake. That's no longer the case, thanks to advanced image generators like Google’s Nano Banana, which can create highly realistic fake pictures. To help prevent its technology from being misused, Google developed an identification tool called SynthID.
SynthID works as a digital watermark integrated directly into photos produced by Nano Banana. This watermark is invisible to the naked eye but acts as a unique fingerprint, allowing users to recognize images generated by this AI. Gemini, Google’s AI tool, already includes a feature to identify AI-generated images using SynthID, and Google announced at its recent I/O conference that this same capability would be extended to more of its products.
Google Search, Lens, and Circle to Search help spot fakes
You can now use Google Search, Google Lens, or the Circle to Search feature to ask if a photo is authentic or if it was generated by AI. Just submit the image, and Google tries to determine its origins.
In regions where “Gemini in Chrome” is available, the ability to detect AI-generated images is being integrated into the browser. That could make checking the authenticity of an image as simple as a browser click.
Beyond Google: SynthID goes multiplatform
This detection isn’t limited to Google’s own images. According to Google, SynthID’s invisible digital marker is being rolled out with other companies, including ChatGPT. “With digital media on so many platforms, it's essential for the entire industry to collaborate and use trustworthy, interoperable tools. More and more online content will carry these imperceptible watermarks, as companies such as OpenAI, Kakao, and ElevenLabs integrate SynthID into a growing share of their AI-generated content,” the company said.
SynthID is becoming ever more important as web disinformation rises.
SynthID’s massive scale
On Gemini, the AI-image detection technology based on SynthID has already been used over 50 billion times. Google reports that, as of now, SynthID has applied its digital watermark to over 100 billion images and videos, and to audio equivalent to more than 60,000 years in duration.
Next time you see a photo that looks a little too good to be true, Google’s AI might already have flagged it.
🟣 For tech news updates, follow Journal du Geek on Google and WhatsApp. They also offer a morning newsletter.
English (US) · 
