Google pulls Google Earth AI image tool a day after launch over fakes

Google is pulling the AI image-generation feature it added to Google Earth on July 30, roughly a day after it went live. The company said it is rolling the feature back while it works on ‘stronger guardrails,’ after researchers and reporters showed the tool could fabricate convincing fake images of disasters, fires and nuclear sites at real map coordinates.

The tool, built on Google’s Nano Banana image model, let users of Google Earth’s web version zoom to any coordinate and generate a photorealistic image layered over real satellite, aerial and 3D data. That combination — synthetic imagery pinned to a real location — is what alarmed critics, who warned it could seed geolocated fakes of events that never happened.

In an update to its original announcement, Google said it was ‘rolling back this feature in Google Earth while we work on implementing stronger guardrails,’ citing ‘screenshots of generated imagery that appear to violate our policies.’ The company said generated images carried a SynthID watermark and did not appear in other users’ views by default.

Those safeguards did not settle the core concern. A watermark travels with the file, but a screenshot of the image, stripped of metadata, does not. Google did not say when the feature would return or in what revised form, and did not disclose how many images were generated before the rollback.

The retreat lands as regulators and platforms grapple with AI-generated media attached to real-world places and people. Google shipped the feature, saw it used to make plausible fakes at real coordinates within a day, and withdrew it on the same day the abuse was documented. The fast reversal shows how little runway a consumer AI tool now gets before misuse forces a rethink.