Google pulled an AI image-generation feature from Google Earth's web version on July 31, roughly one day after launching it, following warnings from researchers and journalists that the tool could be used to fabricate convincing fake satellite imagery of real-world events.
The feature, built on Google's Nano Banana model, let users zoom into any location on the map, click “create image,” and generate a synthetic scene from a text prompt overlaid on real geographic coordinates. Within a day of launch, researchers had used it to produce a fabricated blast crater in Los Angeles, invented protesters outside Google's Mountain View campus, a staged fire engulfing Iran's Kharg Island, a flooded U.S. Capitol, refugees at the Mexican border, and a fictitious nuclear plant in Iran.
Researchers Sound the Alarm
The fabricated imagery was flagged by tech outlet 404 Media and NPR, along with open-source researcher Henk van Ess and Bellingcat researcher Jake Godin — investigators who rely on Google Earth as a trusted reference point when verifying breaking news and documenting atrocities. Their concern was that AI-generated scenes rendered directly onto real map coordinates could be mistaken for authentic satellite capture, undermining a tool widely used for open-source verification work.
Google initially defended the feature by noting that every generated image carries a SynthID watermark identifying it as AI-generated in tools like Gemini, and that the system blocks image creation on certain harmful topics. That defense did not stop the company from reversing course within roughly 24 hours of launch.
Why Verification Tools Matter to the Crypto Industry
Google said in a statement that “we know that people uniquely trust Google Earth for a reliable view of the world,” acknowledging that the incident cut against the platform's core credibility. The company said it is rolling back the feature while it builds stronger guardrails, though it has not given a timeline for reinstating any version of the tool.
The episode is a reminder of why content-provenance efforts — including blockchain-based approaches to timestamping and verifying the origin of images and video — have drawn growing interest as generative AI tools make convincing fabrications easier to produce. Systems that anchor a media file's origin and edit history on a public, tamper-resistant ledger are increasingly pitched as a complement to watermarking schemes like SynthID, precisely because watermarks can be stripped or ignored while an on-chain provenance record is harder to quietly erase.
Google has not said whether it will incorporate any third-party or blockchain-based provenance standards into a future version of the Earth AI tool, but the rapid about-face illustrates how quickly trust can erode once a mapping platform used for real-world verification starts generating convincing fiction.
