Google Earth AI tool lasted one day

Google launched a tool Thursday that let you edit satellite photos with text prompts. You typed what you wanted, and the AI changed the image.

It worked for about one day. Then Google shut it down.

The tool used a model called Nano Banana 2. You could turn a street into a beach or add anything to the sky. It was basically a deepfake generator for the real world.

Digital Digging's Henk van Ess tested it. He added refugees near the Mexican border. He added a bomb crater by a hospital in Gaza. He was testing the limits, not making propaganda.

Google's first response said the images got a digital watermark. They said they blocked "image creation on harmful topics." That's the problem. Who decides what's harmful? You can't ask the AI to judge its own output.

Then Google changed its statement. The tool was gone.

The whole experiment lasted less than 24 hours. A tool that could reshape reality for anyone with a keyboard was pulled before most people even heard about it.

This is how these tools move. They launch, someone pushes them, the company panics, and the tool disappears. No transparency. No process. Just a feature that existed and then didn't.

The people who got burned by this — the communities that were represented or misrepresented in those AI-generated images — they don't get a say. The tool was built, tested, and killed before most Brown people even knew it existed.

Why this matters for us: When a tool that can alter the reality of our neighborhoods gets pulled without warning, our communities lose the chance to push back or shape how these systems represent us.