Google has already disabled a newly introduced AI image feature in Google Earth after just 24 hours. With the model Nano Banana 2, it was possible to generate fictitious scenes from real satellite data since July 30, 2026, such as an explosion in Paris or a supposed nuclear facility in Iran. Google responded to criticism from geodata experts and temporarily withdrew the feature.
Fake images show explosions and alleged military facilities
The feature named “Create Image” could be used in the web version of Google Earth. Users selected any location on the map, entered a text prompt, and the model generated a photorealistic image based on the real satellite, aerial, and 3D data of that point.
According to reports from NPR, test users created, among other things, an explosion in Paris, a supposed nuclear facility in Iran, a bomb crater in Russia, and a training camp of the terrorist militia IS in Syria. Invented scenes such as floods in Bangladesh could also be depicted. Geodata experts warned that this could create credible-looking evidence for events that never occurred.
Google announces stricter protective measures
Google reacted within 24 hours and disabled the feature again on July 31, 2026. A company spokesperson stated according to TechCrunch, that while they had observed meaningful use by geodata experts, they also saw shared screenshots that violated their own guidelines. Google wants to use the time to implement stronger protective mechanisms before the feature returns.
Technically, “Create Image” is based on the image model Nano Banana 2, internally also referred to as Gemini 3.1 Flash Image. As beckmann.ai previously reported, a lighter variant of the model was already in use in Google Images’ AI overviews in mid-July. The current pause only affects the browser version of Google Earth; the mobile apps never received the image generation feature.
Criticism follows a familiar pattern with AI image generators
The quick withdrawal is part of a series of similar cases. Just in mid-July, Meta withdrew its image feature Muse within a few days after criticism from users and talent agencies, after other people’s Instagram photos could be used for AI images without consent. Both cases share a common pattern: the companies first roll out new generative features widely and only make improvements after public criticism, rather than testing for abuse scenarios in advance.
In the case of Google Earth, the photorealism feature was tied to a product that users traditionally perceive as a reliable representation of reality. It will be crucial what specific protective mechanisms Google implements before a relaunch, and whether restrictions on security-sensitive locations such as military facilities can be reliably enforced. The company has not yet given a timeline for a return.
