Google pulls Earth's AI image tool after disinformation fears

On July 30, 2026, Google introduced a new 'create image' feature in Google Earth, leveraging its Nano Banana 2 model to let users generate custom AI visualizations directly on satellite imagery. The tool was positioned as a creative extension of Google's geospatial platform, allowing individuals to overlay synthetic landscapes, buildings, or environmental changes onto real-world locations. However, the feature barely survived a day. Within less than 24 hours, Google rolled it back, citing concerns raised by researchers and open-source intelligence (OSINT) experts. These experts quickly identified the potential for the tool to create convincing but entirely fabricated satellite images, which could be used to spread disinformation about troop movements, disaster damage, or infrastructure projects. The speed of the rollback suggests Google recognized the severity of the risk, but the company has not yet clarified whether the feature will return in a more constrained form.

The controversy sits at the heart of a growing challenge for AI companies: generative models are becoming so good at producing realistic imagery that distinguishing fact from fabrication is increasingly difficult. While Google has introduced labels and watermarks for AI-generated content, satellite imagery poses unique problems. Geo-located fakes are particularly dangerous because they can be shared as 'proof' of events that never occurred, influencing public opinion and international policy. This incident also mirrors broader industry debates about responsible AI deployment, where generative capabilities race ahead of safeguards. For journalists, humanitarian organizations, and defense analysts, the episode is a warning that verification of satellite imagery, once considered inherently trustworthy, is now as essential as verifying text or social media posts. Google's rapid decision to pull the feature may set a precedent for how other companies handle geospatial AI tools.