Google has removed a new AI feature from its Google Earth platform that enabled users to generate artificial satellite images using text prompts. The decision came swiftly after the feature's release, following widespread concern from experts and open-source investigators who highlighted its potential for creating and spreading misinformation.
The feature, which allowed users to "zoom in to a place in Google Earth on web, tap 'create image,' and type whatever you want to see," was initially intended for useful purposes by geospatial professionals. However, issues arose immediately, with examples like fake nuclear power plants in Iran, refugee camps on the US-Mexico border, and bomb-damaged hospitals in Gaza being easily generated. Critics, such as misinformation expert Henk van Ess, demonstrated how this tool could be used to weld invented imagery onto genuine coordinates, undermining the credibility historically associated with satellite data.
Despite Google's initial assertion that all AI-generated images would be watermarked with SynthID and detectable by Gemini or Google Lens, these protections were found to be insufficient. Tests showed that Gemini could be tricked into confirming fake images as real, and some users noted that many people would not think to verify images using Google's tools. The company acknowledged that people uniquely trust Google Earth for a reliable view of the world and stated it would implement stronger guardrails before redeploying the feature.
The removal underscores the challenges of integrating generative AI with platforms relied upon for factual information. Experts like Bill Greer, a geospatial analyst, noted that satellite imagery has traditionally been a trustworthy data source, making the potential for AI-generated deepfakes particularly damaging to public trust. The incident highlights the need for robust safeguards and careful consideration of how AI tools can be misused, especially in sensitive areas like geopolitical analysis and breaking news verification.
