Google withdraws AI satellite image tool after disinformation concerns

Google has paused a Google Earth feature that let users generate artificial satellite images from text prompts, following strong criticism from journalists, human rights advocates and open-source investigators.

Geoff Brumfiel reports for NPR that Google removed the tool one day after introducing it. The company says it saw screenshots of generated images that appeared to breach its policies and is now working on stronger safeguards.

The feature allowed users of Google Earth on the web to select a location, choose “create image” and describe what they wanted to see. Google presented it as a creative tool, but analysts warned that it could make fabricated scenes look like credible evidence from space.

Open-source researcher Henk van Ess told NPR that prompts involving refugees at the Mexican border, an Iranian nuclear plant, an Amsterdam crash and bomb damage at a Gaza hospital were not rejected. NPR also generated images showing a fire on Iran’s Kharg Island and flooding at the U.S. Capitol, despite neither event having occurred.

Why satellite imagery matters

Satellite imagery is widely used to verify events in areas that reporters and investigators cannot reach. Analysts compare images from space with videos, photographs and other material to establish where and when an event happened.

Jake Godin of Bellingcat told NPR that satellite images have long been viewed as relatively difficult to fake because they originate from specialist cameras and providers. A simple image-generation button could lower the effort needed to create convincing false claims, he said.

Google says generated images carry its SynthID watermark and that it blocks image creation on harmful subjects. Critics argue that visible or detectable labels do not solve the wider problem: fabricated imagery can spread quickly, while governments and other actors may dismiss authentic satellite evidence as AI-generated.