Google integrated its Nano Banana 2 [1] AI image generator into Google Earth on July 31, 2026 [2], allowing users to place AI-generated objects onto satellite imagery.
This update creates a significant risk for the spread of deceptive content because it blends synthetic images with trusted, real-world geographical data. Analysts said that the ability to manipulate satellite views could be used to manufacture evidence of events that never occurred.
Experts from BBC Verify and other analysts said the tool could be misused to create convincing fake images. The primary concern is that invisible watermarks intended to identify AI content may not be sufficient to stop the spread of misinformation [2].
"Google's new AI feature could be weaponised to produce false satellite imagery that looks authentic," Dr. Emma Tucker said [2].
To demonstrate the capability of the tool, reports highlighted two examples of manipulations [1]. These included an image of a collapsed Eiffel Tower, and another showing a sinkhole swallowing the Great Pyramid of Giza [1].
Critics argue that the technical safeguards are inadequate. John Smith said that invisible watermarks are not foolproof, and that third-party detection tools can be tricked into labeling manipulated Google Earth images as real [3].
Google has not provided a detailed response to these specific concerns regarding the bypass of its watermarking system, though the tool is now active on the platform [2].
“Google's new AI feature could be weaponised to produce false satellite imagery that looks authentic.”
The integration of generative AI into geospatial data represents a shift from AI creating standalone images to AI altering verified records of the physical world. If detection tools cannot reliably distinguish between a real satellite capture and a Nano Banana 2 modification, the utility of satellite imagery as a source of objective truth in journalism and intelligence may be diminished.
