Via theverge.com
The Nano Banana feature in Google Earth let users generate AI images over real satellite maps, and the misuse concerns wrote themselves.
Google launched an AI-powered image generation tool inside Google Earth on July 30, 2026. By July 31, it was gone. The feature, called Nano Banana, let users overlay AI-generated visuals onto real satellite and aerial imagery using simple text prompts. The internet immediately figured out the worst possible uses for it, and Google hit the kill switch.
What Nano Banana actually did #
The tool was designed to let users generate custom AI images that would sit on top of real-world satellite, aerial, and 3D imagery inside Google Earth. You’d type a text prompt, and the system would produce a visual overlay, essentially letting you paint AI-generated content onto the actual planet.
Google pitched it as a creative and educational tool. Think: visualizing historical events at their real locations, mocking up real estate concepts on actual plots of land, or designing futuristic cityscapes over existing skylines.
The company also built in guardrails. Generated images were watermarked as AI creations. They were kept separate from the core Google Earth experience, meaning other users browsing the map wouldn’t stumble across someone’s AI fantasy layered onto real geography. The images lived in a sandbox, visible only to the person who made them.
The backlash was immediate and specific #
Within hours of launch, critics highlighted that Nano Banana could be used to generate fake satellite imagery depicting drone strikes, altered national borders, or fabricated military installations, all rendered on top of real, recognizable geography.
Satellite imagery carries a particular kind of authority. Journalists, researchers, and intelligence analysts have long relied on Google Earth as a credible, neutral source of geospatial data. A tool that makes it trivially easy to generate convincing fakes of that data, even with watermarks, threatens to erode that trust.
Google pulled the feature one day after launch. The company emphasized the watermarking and sandboxing protections it had built in, but ultimately decided the reputational risk wasn’t worth the engagement upside.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our