Google rolls back AI image generator in Google Earth amid misinformation concerns
Google has rolled back a feature that let users generate satellite images with AI after researchers demonstrated how it could spread misinformation.
Feature and its misuse
Google Earth recently introduced a feature called "Nano Banana 2" on its web version, allowing users to generate images from satellite, aerial, and 3D mapping data using text prompts. Initially pitched as a tool for visualizing historical sites and real estate projects, it quickly drew attention when researcher Henk van Ess from Digital Digging demonstrated how it could create misleading images, such as "refugees near the Mexican border" and a "bomb crater near a hospital in Gaza."
Google's response and verification methods
Google responded by emphasizing that it takes misinformation seriously and that every image generated with the feature includes a SynthID digital watermark. Users were advised to check suspicious images through the Gemini app or Google Lens in Search. Van Ess also suggested comparing images with other satellite platforms like Sentinel-2 or Landsat, and verifying orbital data such as the capture time and satellite name.
The risk of misleading images and the rollback
While some generated images were easily recognizable as fake—like the Sphinx merged with the Statue of Liberty—others could be convincing enough to spread misinformation. Digital Digging even managed to fool Hive's AI detector with an AI-altered video from Google Earth, showing that the watermark can be circumvented or missed. Following these concerns, Google announced on July 31st that the feature has been rolled back.
