Google has removed a generative AI feature from Google Earth after users created and shared realistic-looking satellite images that were not genuine. The feature was reportedly available for less than 24 hours before being taken down.
The controversy began when AI-generated images started circulating online, including supposed scenes showing refugees at the US-Mexico border, a burning nuclear facility in Iran, a plane crash in Amsterdam, bombed residential areas and oil refinery fires. BBC Verify and other fact-checkers examined some of the viral content and determined that the images were AI-generated rather than authentic satellite imagery.
The incident has raised serious questions about the use of generative AI in mapping and other platforms that people traditionally associate with reliable geographic information.
Google reportedly removed the “Nano Banana 2” feature and said it would introduce stronger safeguards before making similar tools available again. The company also plans to add AI watermarks to newly generated images to help users identify synthetic content.
Google’s mass layoffs: Hundreds of Employees Terminated Overnight
However, the episode highlights a larger challenge: AI-generated images can look increasingly convincing, making it difficult for ordinary users to distinguish real satellite imagery from fabricated scenes. During wars, natural disasters or other major emergencies, misleading visuals can spread rapidly and potentially create confusion or panic.
Satellite imagery is often used by journalists, researchers, governments and humanitarian organizations to verify events happening on the ground. If fabricated images are presented as genuine evidence, they can undermine public trust in authentic imagery.
The incident also raises an important question about responsibility. Should powerful AI image-generation tools be available inside platforms that users rely on for factual geographic information?
Google’s decision to remove the feature suggests that stronger safeguards may be necessary before AI-generated content is integrated into public mapping services.
Should AI be allowed to generate or edit satellite-style imagery on public mapping platforms?
