Google's AI Image Generator Is Impressive, Unhinged, and Coming for Your Photo Albums

Google recently dropped Nano Banana 2, the updated version of its AI image generator now baked into Gemini, and the results are equal parts impressive and deeply strange. It's faster than its predecessor, pulls real-time data from the web, and can edit existing photos with a level of detail that would have seemed impossible two years ago. It also made a writer look 80 years old when all they asked for was hot tub wrinkles, which honestly feels like a metaphor for AI in general.

The capabilities are genuinely worth paying attention to. Nano Banana 2 can generate infographics pulled from live web data, recreate intricate details from uploaded photos including jewelry on an underwater hand, and composite a person's face into entirely fabricated scenes with photorealistic backgrounds. The good news is that Google watermarks its AI-generated outputs. The sobering news is that those watermarks are easy to miss while scrolling, and the outputs are improving fast enough that "easy to miss" is doing a lot of heavy lifting in that sentence.

For IT and security professionals, this is a content authenticity conversation, not just a fun tech demo. The same tool that put a writer in a snow-covered hot tub can put anyone anywhere, doing anything, in seconds. Enterprise security awareness training that doesn't include a module on AI-generated image manipulation at this point is running behind the threat landscape. Deepfakes used to require skill and time. Nano Banana 2 requires a Gemini account and a creative prompt.

The photo editing capabilities are where things get genuinely interesting from a governance standpoint. This tool can take a real photo of a real person and place them in a completely fabricated context convincingly enough to fool a casual viewer. That's not a hypothetical risk. That's a feature available for free in the Gemini app right now. The policy conversations organizations need to be having about AI-generated content aren't coming. They arrived while everyone was watching the skiing demo.

The practical upside is real too. Marketers, content creators, and communications teams have access to a tool that can dramatically accelerate visual content production. IT leaders rolling out Gemini for enterprise use should be thinking about acceptable use policies for image generation alongside the more obvious concerns about data privacy and hallucinated outputs. Nano Banana 2 is not the last of its kind. It's just the latest reminder that the visual information environment is changing faster than most organizations' governance frameworks.

https://www.wired.com/story/google-nano-banana-2-ai-image-generator-hands-on/