Google Pulls AI Image Tool From Google Earth After Backlash

Google Pulls AI Image Tool From Google Earth

The feature, which was initially unveiled on Thursday, was powered by Google’s Nano Banana 2 image generation model. It allowed users to generate photorealistic images grounded in Google Earth’s satellite, aerial, and 3D mapping data simply by typing a text prompt for any location on the globe. Google suggested in its original announcement that the capability could be used to bring history to life—such as rendering Pompeii ruins as they looked in 78 A.D.—or to create real estate plans and custom infographics.

Misinformation Concerns and Dangerous Deepfakes

Within hours of the launch, however, social media flooded with alarming AI-generated images created using the tool. Open-source intelligence (OSINT) experts and security researchers demonstrated how easily the feature could be manipulated to fabricate realistic scenes of disasters, conflicts, and sensitive sites.

Online research methods expert Henk van Ess used the tool to create deepfakes including fake refugee camps along the US-Mexico border and a nuclear plant in Iran. Other users generated imagery depicting a nonexistent flood, a 9/11-style attack, a crater from an explosion near a hospital in Gaza, and a large-scale accident on the streets of Amsterdam.

Experts warned that embedding a generative tool directly into a trusted verification platform posed severe risks. Sam Stockwell, a senior fellow at the Alan Turing Institute, noted that Google Earth has served for decades as a reference layer for journalists and investigators to verify suspicious imagery from hard-to-reach regions, and embedding generative AI inside it undermines that measure. Brady Africk, an analyst at the American Enterprise Institute, added that because the fakes were tied to real satellite imagery, they eliminated common AI artifacts like terrain or scale mismatches.

Guardrails, Watermarks, and Official Alibis

Google stated that the generated images were watermarked as AI-generated and did not appear in the main public Google Earth experience. However, researchers argued that these safeguards were insufficient. Emmanuelle Saliba, chief investigative officer at AI detection company GetReal, told Jpost that verification typically requires additional effort while images are consumed and shared in seconds. Henk van Ess also noted that watermarks only worked using specific Google products, such as Gemini, and warned of another danger: authorities could potentially use the existence of the tool to claim true photographs of atrocities are falsehoods.

By Friday, Google reversed course and paused the capability. A Google spokesperson stated that while geospatial professionals used the feature for useful purposes, the company observed users sharing screenshots that appeared to violate policies. Google announced it was temporarily disabling the feature while working on stronger guardrails.