Google has abruptly pulled its new Google Earth AI tool less than 24 hours after launch, following widespread concerns that the feature could produce and spread misleading fake imagery. The tool, which used the company’s Nano Banana 2 model to overlay AI-generated images onto real satellite data, was removed after researchers and users reported screenshots that appeared to violate Google’s content policies.
Launched on July 30 via Google Earth’s web platform, the feature allowed users to enter text prompts and generate new scenes using satellite, aerial, and 3D mapping data. However, within hours, open-source intelligence researchers demonstrated how the tool could create images of sensitive locations, raising red flags about potential misinformation. Researcher Henk van Ess was among those who tested various prompts, showing that the generated content could easily be misused.
In a statement, Google acknowledged the issue: “We know that people uniquely trust Google Earth for a reliable view of the world. We’ve also seen people sharing screenshots of generated imagery that appear to violate our policies.” The company emphasized that the generated images did not alter the main Google Earth view and that each AI-created image carried a SynthID watermark to help identify synthetic content. Additionally, Google said Gemini and Google Lens could help users detect AI-generated images.
Despite these safeguards, experts warned that screenshots of the generated visuals could still spread misleading information across social media, bypassing the watermarks. The quick rollback underscores a growing challenge for AI companies: balancing innovation with the risk of amplifying misinformation. Google Earth holds strong public trust as a reliable source of real-world imagery, and the introduction of AI-generated scenes threatened to undermine that credibility.
Google now plans to strengthen its safety controls before deciding whether to restore the feature. The incident highlights how rapidly AI products can face safety concerns once exposed to public testing. As AI generation tools become more common, companies will need to implement robust safeguards to prevent misuse while still delivering innovative user experiences.
