Google Earth’s AI deepfake tool only lasted one day
Google launched and then quietly removed an AI-powered image-editing feature in Google Earth within a single day. The tool, which used a model referred to as Nano Banana 2, allowed users to modify satellite imagery using text prompts - effectively letting anyone overlay fictional or fabricated elements onto real geographic locations.
The feature attracted immediate scrutiny from researchers and journalists. Henk van Ess of Digital Digging demonstrated the tool's potential for misuse by generating images that placed refugees near the US-Mexico border and showed a bomb crater adjacent to a hospital in Gaza. These examples were deliberately chosen to illustrate how the tool could be used to produce convincing but entirely false depictions of real places, with the added credibility that comes from satellite-style imagery.
Google's initial response emphasized that generated images carried a digital watermark and that the system had guardrails against "harmful topics." However, those protections did not appear to cover the range of scenarios critics raised. By Friday, Google had updated its position and shut the feature down entirely - a notably swift reversal for a product that had only just been announced.
The episode highlights a recurring tension in the deployment of generative image tools: the gap between what a content policy says on paper and what users can actually produce in practice. Satellite and geographic imagery carries a particular weight, since it is commonly associated with documentation and evidence. Adding AI generation on top of that creates a format that can look authoritative while depicting something that never existed. Google has not said whether a modified version of the feature will return.
