Google’s Google Earth AI Dream Collided With Deepfake Reality in 24 Hours

Google’s Google Earth AI Dream Collided With Deepfake Reality in 24 Hours
Google’s latest AI flourish lasted barely a day. The company rolled back a new Google Earth feature after it became obvious that a tool meant to reimagine places could just as easily fabricate convincing scenes from the real world.

The pitch was straightforward: bring Nano Banana 2 into Google Earth so users could remix satellite, aerial, and 3D imagery with text prompts, whether to visualize a renovation, a historical reconstruction, or a community project. Google framed it as a creative and practical upgrade to a trusted mapping product, with one promotional post asking, “What if you could see a city as it looked 100 years ago, or visualize a new basketball court at your local community center?”

But the backlash landed just as fast as the launch. Critics warned that dropping an image generator into a platform associated with factual, geographic reality was an open invitation to misinformation. One report summed up the concern bluntly: “Here’s the problem with putting an AI image generator in Google Earth.” Testing described by The Verge and researcher Henk van Ess showed the tool could generate inflammatory fake scenes, including refugees near the Mexican border and a bomb crater by a Gaza hospital, while still looking plausible enough to circulate.

Google initially leaned on safeguards, saying generated images carried SynthID watermarking and did not appear in the main Google Earth experience for others to see. But that defense looked thin once examples surfaced suggesting watermarks and detection tools were not enough. As one account of the rollout put it, the feature effectively let users create “AI deepfakes of the real world.”

By Friday, Google had reversed course, saying it was “rolling back this feature in Google Earth while we work on implementing stronger guardrails.” That retreat captures the real tension here: AI companies want their tools embedded in everyday products, but the closer those products sit to the public record, the smaller the margin for error becomes.

Continue reading https://foxvector.com/stories/019fba0e-7d3f-102d-738e-0023b78a0355