Google removed an AI-powered image generation feature from Google Earth after researchers, journalists, and open-source intelligence analysts demonstrated it could easily produce misleading satellite imagery of fabricated events.
The tool, which used Google’s Nano Banana 2 image generator, launched on July 30 and was pulled roughly a day later after a wave of criticism over its potential for misuse.
The feature allowed anyone to zoom into a location in Google Earth’s web version, tap “create image,” and type a prompt to generate AI imagery using real satellite photos as a base. Within hours, researchers showed the tool would produce photorealistic depictions of events that never happened.
Dutch open-source intelligence researcher Henk van Ess told NPR he tested the tool extensively: “I tried refugees at the Mexican border, a nuclear plant in Iran, a crash in Amsterdam, a hospital with a bomb crater in Gaza. Nothing was refused”. NPR itself fabricated an image of Iran’s Kharg Island on fire, while the BBC produced images of a collapsed Eiffel Tower and generated fake damage at the Great Pyramid of Giza.
Google acknowledged the problem in a statement on X: “We’ve seen geospatial professionals using this feature for a range of useful purposes, however we’ve also seen people sharing screenshots of generated imagery that appear to violate our policies. So we’re rolling back this feature in Google Earth while we work on implementing stronger guardrails”.
Threat to Verification
For the open-source intelligence community, the concern runs deeper than a single botched product launch. Satellite imagery has served as a bedrock of verification work, used to confirm events on the ground and debunk suspect footage. Jake Godin, a senior researcher with the investigative group Bellingcat, told NPR that satellite images have historically been “a safe bet when it comes to verifying an event because it’s hard to fake”.
The integration did “immediate, catastrophic damage to the credibility of satellite imagery,” the account OSINTtechnical, which has nearly 2 million followers on X, said in a post.
Watermarks Offered Little Protection
Google pointed to its SynthID watermark system and the fact that generated images did not appear in the main Google Earth experience. But researchers found that the invisible watermarks were not reliably detectable even by Google’s own Gemini AI, and users could simply screenshot images and share them without any visible provenance signal.
For van Ess, the episode struck at the heart of the problem: the tool meant “the fake is made inside the thing people use to check whether pictures are true”.
Leave a Reply