Google Earth AI Tool Pulled After Users Deepfaked Disaster Scenes

Google pulled a new AI image-editing feature from Google Earth less than 24 hours after launch. The tool allowed Gemini’s Nano Banana 2 image model to access Google Earth and transform locations using text prompts. The rapid rollback shows how quickly generative AI features tied to real-world geography can create trust and safety problems.

Users created realistic-looking deepfakes depicting disasters and sensitive scenarios. AI researcher Henk van Ess created fake refugee camps along the US-Mexico border and a nuclear plant in Iran, according to the article. Other users generated images of a nonexistent flood and a 9/11-style attack, highlighting the risk of misleading screenshots spreading outside their original context.

Google said geospatial professionals had used the feature for useful purposes, but that some shared generated imagery appeared to violate company policies. A Google spokesperson said the company was rolling back the feature in Google Earth while it worked on stronger guardrails. Google also said the AI-generated images were watermarked and never appeared in the public Google Earth experience.

The episode adds to a pattern of AI tools being released and then quickly withdrawn after unintended uses emerge. Business Insider noted Meta’s Muse Image feature, launched on July 7, was removed in less than a week after the company said it had “missed the mark.” For readers, the practical takeaway is clear: treat viral AI-generated location images with caution, especially when they depict emergencies, conflict, or infrastructure.