Google has rolled back a newly launched artificial intelligence feature in Google Earth after researchers demonstrated its potential to generate realistic fake imagery, according to reporting by Ars Technica.
The feature, known as Nano Banana 2, allowed users to create altered images using text prompts applied to satellite, aerial, and 3D imagery. Google originally launched the tool globally on the web version of Google Earth earlier in the week, positioning it as a way to visualize historical sites and real estate projects.
Henk van Ess of Digital Digging demonstrated how the tool could be misused by generating imagery showing refugees near the Mexican border and a bomb crater near a hospital in Gaza. Another generated image combined the Sphinx with the Statue of Liberty.
Watermarks and Detection Bypasses
Google initially responded to the concerns by pointing to its SynthID digital watermark embedded in every image created with Nano Banana in Google Earth. The company stated that users could check whether an image was synthetic by using Lens in Search or by asking the Gemini app.
However, Digital Digging successfully bypassed AI detection systems, fooling Hive's AI detector with an AI-altered video created from Google Earth imagery. Although Google's digital watermarks are difficult to remove, researchers showed they remain vulnerable to being missed or bypassed.
To verify suspicious imagery, van Ess recommended comparing images to alternative satellite platforms such as Sentinel-2 or Landsat. He also suggested using Google's "@verifyai" tag in Gemini and verifying orbital data, including when the image was taken and by which satellite, to confirm authenticity.
