Google Pulls Earth AI Feature After Deepfake Satellite Image Concerns
Search giant removed Nano Banana 2-powered tool less than 24 hours after launch following demonstrations that it could fabricate convincing fake imagery of sensitive sites.
Google launched and killed an AI image generation feature in Google Earth within 24 hours, removing the tool on July 31, 2026, after researchers demonstrated it could create convincing fake satellite imagery of real-world locations.
The feature, powered by Google's Nano Banana 2 model (also identified as Gemini 3.1 Flash Image), went live July 30, 2026. It allowed users to generate AI imagery layered over authentic satellite, aerial, and 3D map data on Google Earth's web and desktop platforms.
Google removed the feature after journalists and investigators showed it could fabricate imagery tied to genuine coordinates, inheriting the credibility of the platform's mapping data. The company cited a need for stronger guardrails after seeing generated imagery that violated company policies.
"We know that people uniquely trust Google Earth for a reliable view of the world," Google said in a statement posted on X. "We've also seen people sharing screenshots of generated imagery that appear to violate our policies."
Investigator Henk van Ess demonstrated the vulnerability by creating fake images including a nuclear plant in Iran. "Google spent twenty years building the reference the world checks against," van Ess said. "Today it added a button that makes things up."
Google defended the feature by noting it used SynthID digital watermarks to identify AI-generated content. However, BBC Verify found it was possible to trick the Gemini chatbot into claiming fake Google Earth images were real, effectively bypassing the SynthID checks.
The incident highlights risks as Google integrates generative AI across its core products. Satellite imagery has long been treated as an unblemished source of truth for journalists, researchers, and intelligence analysts. The ability to create fake imagery of conflict zones, nuclear facilities, or disaster areas tied to real coordinates poses significant risks to global security and journalistic integrity.
Google Earth has served as a gold standard for reliable geospatial data since its launch. By allowing users to invent imagery anchored to genuine locations, the company introduced a mechanism where fabricated content could inherit the platform's established credibility.
The rollback comes amid broader criticism of Google's AI integrations, including accuracy issues and hallucinations in its Gemini chatbot. The company has not announced whether it plans to reintroduce the feature with additional safeguards.