IC Technology & Cyber — 2026-08-02

Google Earth AI Feature Enables Fabrication of Realistic Satellite Imagery Threatening OSINT and GEOINT Analysis

BLUFEven with the rollback, satellite imagery's evidentiary standing is durably compromised, and a Google relaunch with safeguards adequate to restore OSINT confidence remains unlikely within three months.

Google added an AI image-generation feature to Google Earth's web browser version on July 30, powered by its Nano Banana 2 model, allowing users to fabricate photorealistic satellite imagery anchored to real coordinates 1. Within hours, researchers including Dutch OSINT analyst Henk van Ess, NPR, BBC Verify, AFP, and the Federation of American Scientists generated fake imagery of a nuclear facility in Iran, fires at Iran's Kharg Island oil terminal, flooding at the US Capitol, war damage in Gaza and Kyiv, and a fabricated missile silo, all rendered on authentic Google Earth data 12. Google's stated safeguard, the SynthID watermark, failed to reliably flag the fakes: Golden Owl researcher Tal Hagin and Futurism both tested a fabricated Gaza hospital image through Google's Gemini chatbot and received responses stating no reliable AI-generated signal could be detected 13. Google rolled back the feature on July 31, less than 24 hours after launch, stating it would implement stronger guardrails before any relaunch and noting the generated images had not appeared in the main Google Earth experience for other users 134.

Analysis
The rollback removes the tool, not the demonstrated capability: satellite imagery's credibility as a verification baseline is now durably degraded, giving state and non-state actors plausible grounds to dismiss authentic imagery as fabricated. Relaunch with guardrails sufficient to satisfy OSINT critics is unlikely within the next three months, since SynthID's core flaw, a passive watermark requiring an extra verification step most viewers never take, is a design problem no policy adjustment alone resolves. Moderate confidence reflects Google's stated intent to rebuild rather than abandon the feature, set against the absence of any disclosed technical fix or relaunch timeline.
4 sources
  1. Google Earth AI Pulled After Fakes of Nuclear Facilities Pass Watermark Check - Tech Times
  2. Google Earth New AI Lets Anyone Fabricate Completely Bullshit Satellite Images - 404 Media
  3. Google Pulls Down Feature for AI-Generating Fake Satellite Images on Google Earth After It Immediately Turns Into a Blatant Misinformation Crisis - Futurism
  4. Google pauses AI satellite images, after fears of deepfakes in the sky - NPR

View in full brief →

UNCLASSIFIED // OPEN SOURCE