Google yanked its Earth AI in 48 hours after users faked bombed hospitals

Google launched and killed a feature called Nano Banana inside Google Earth in under 48 hours. The idea seemed playful enough: use Nano Banana 2 to generate custom scenes from satellite imagery. Turn the ruins of Pompeii into a bustling Roman street. Drop a planned building onto an empty lot. Harmless, even useful, in the right hands.

It didn’t stay in the right hands for long. Users immediately generated convincing fake satellite images of a refugee column at the Mexican border and a bombed-out hospital in Gaza. Google insists the images were labeled as AI-generated and invisible to other users within Earth, but that didn’t stop people from taking screenshots and sharing them widely. Those shares violated Google’s policies, and the company killed the feature. “We know that people uniquely trust Google Earth for a reliable view of the world,” Google said in a statement. No timeline for a return—only a promise that it won’t come back until stronger safeguards are in place.

The speed of the rollout and reversal is jarring, even by Google’s standards. The company has fumbled AI launches before, but handing users a frictionless tool to rewrite reality inside what many treat as a digital atlas feels like a different order of mistake. It’s one thing for a generative image model to produce a fake cityscape. It’s another to bake that capability directly into the interface people use to check property lines or verify news footage.

The bigger problem is that pulling the feature doesn’t solve anything. Anyone with Photoshop or a standalone AI image tool can already manipulate a screenshot of Google Earth. What Google did was normalize the idea and remove every possible barrier to entry. It demonstrated, inadvertently, that the line between a satellite photo and a synthetic scene is now a single prompt away—and that genie isn’t going back in the bottle just because one integration got shelved.