Google shut down its AI-powered satellite image editing feature, Nano Banana 2, inside Google Earth roughly one day after launching it. The tool allowed users to modify satellite imagery using simple text prompts, creating deepfakes of real-world locations. Within less than 24 hours, digital ethics researcher Henk van Ess demonstrated the problem by using text prompts to generate images showing refugees near the Mexican border and a bomb crater next to a hospital in Gaza; neither scenario was real, but both looked convincingly like authentic satellite imagery. Google’s initial response cited two safeguards: images generated with Nano Banana 2 included a digital watermark, and the company said it prevented “image creation on harmful topics.” Earlier versions of the Nano Banana model were discussed in Google blog posts as early as February 2026, with the company touting significant improvements in speed, accuracy, and integration with real-time geospatial data. Nano Banana 2 was positioned as a substantial upgrade tied directly to satellite and aerial imagery sources, with the ambition to let users visualize locations historically or futuristically using AI grounded in real geographic data.