Particle.news

Google Pauses Google Earth AI Image Tool After Rapid Abuse

The company says it will add stronger guardrails before restoring the feature because researchers showed it could produce convincing fake satellite scenes tied to real coordinates.

Overview

  • Google introduced a Nano Banana 2–powered “create image” option in Google Earth that let users type a prompt to generate photorealistic overlays on top of satellite, aerial and 3D map imagery.
  • Researchers including Henk van Ess and news agencies used the tool to make fabricated scenes such as a nuclear plant in Iran, refugees at the U.S.-Mexico border and bomb craters, showing the feature could create geopolitically sensitive fakes.
  • Within about 24 hours Google rolled the feature back and said generated images had been watermarked with SynthID and sandboxed, but it will not restore the tool until stronger protections are in place.
  • Verification experts warned SynthID and automated checks are weak when images are screenshotted, filmed from screens or re-encoded, and said easy ‘welding’ of fakes to real coordinates risks eroding trust in satellite imagery.
  • The episode follows a pattern of fast AI rollouts followed by pullbacks and highlights new risks for journalists, investigators and the public who rely on satellite maps as independent evidence.