Google has rolled back a newly launched artificial intelligence feature in Google Earth less than 24 hours after its debut, following concerns that the tool could be used to generate misleading satellite imagery. The company said it had observed users sharing screenshots of AI-generated images that appeared to violate its policies and decided to temporarily withdraw the feature while implementing stronger safeguards.
The feature, introduced on July 30, enabled users to generate AI-enhanced satellite, aerial and 3D map imagery using Google's latest image generation model, Nano Banana 2. The tool was designed to help users visualise concepts such as urban redevelopment, architectural planning, historical reconstructions and landscape transformations by applying text prompts to real-world locations.
However, within hours of the launch, researchers and members of the open-source intelligence (OSINT) community demonstrated how the technology could be misused to create realistic but fabricated satellite images. Examples shared online included fake refugee camps, non-existent military facilities, fabricated disaster zones and altered conflict imagery. Although the generated images did not replace or modify Google Earth's official satellite imagery, screenshots of the AI-generated outputs raised concerns about their potential to spread misinformation if shared without context.
In a statement, Google said it recognised the trust users place in Google Earth as a reliable source of geospatial information. The company noted that while geospatial professionals had already identified useful applications for the feature, it had also seen examples of generated imagery that appeared inconsistent with its policies. As a result, Google decided to pause the rollout until stronger guardrails could be introduced.
Google clarified that AI-generated images created through the feature were watermarked using SynthID and never appeared within the public Google Earth experience viewed by other users. Instead, they remained separate outputs created for individual users. Even so, researchers argued that screenshots could still circulate widely across social media, making it difficult for audiences to distinguish authentic satellite imagery from AI-generated content.
The episode highlights the growing challenges technology companies face as generative AI becomes integrated into products traditionally regarded as trusted sources of information. Google Earth has long been used by journalists, researchers, humanitarian organisations and investigators to verify locations, monitor environmental changes and document events. The introduction of AI-generated imagery into such an ecosystem raised questions about preserving public trust while expanding creative capabilities.
Industry experts have increasingly warned that AI-generated visual content can blur the line between authentic documentation and synthetic media. While watermarking technologies such as SynthID provide one layer of protection, researchers argue that additional safeguards are needed to prevent manipulated images from being taken out of context or repurposed to support false narratives. The rapid spread of AI-generated visuals across digital platforms has intensified calls for stronger governance and clearer disclosure standards.
The rollback also reflects a broader trend among technology companies introducing generative AI features while balancing innovation with safety. Several firms have modified or paused AI products after users identified ways to bypass safeguards or generate harmful content. Companies continue investing in watermarking, content authentication and policy enforcement as AI-generated media becomes more sophisticated.
Google has not indicated when the feature will return but said it is working on additional protections before making it available again. The company maintained that the initial rollout demonstrated valuable use cases for professionals working in mapping, planning and geospatial analysis, while acknowledging that stronger controls are necessary before wider deployment. The incident underscores the increasing importance of trust, transparency and responsible AI development as generative technologies become embedded in widely used digital platforms.