Feature Removed Just One Day After Launch
Google has withdrawn a newly launched artificial intelligence feature from Google Earth following strong criticism from researchers and open-source intelligence specialists.
The company introduced the “Create Image” tool on July 30. It removed the feature one day later after users generated and shared realistic images of events that had never occurred.
The experimental tool allowed Google Earth users to select a real location and modify its appearance through written prompts. It combined Google Earth’s satellite, aerial and three-dimensional mapping data with Google’s Nano Banana 2 image-generation model.
Users could create photorealistic scenes within seconds. They could add buildings, change landscapes or visualise historic locations.
Google initially promoted the feature for education, urban planning and property development. The company said teachers could recreate historical scenes, while architects could visualise projects before construction began.
However, researchers quickly demonstrated that the same technology could be used to create misleading images involving real locations.
Google confirmed that it was rolling back the feature while developing stronger safety measures. The company acknowledged that people place unusual trust in Google Earth as a reliable representation of the physical world.
Researchers Warned of Dangerous Geospatial Disinformation
The feature caused immediate concern among journalists, satellite-imagery analysts and open-source investigators.
Experts warned that realistic images generated over authentic geographical locations could be mistaken for genuine satellite evidence.
Such material could be used to fabricate military activity, natural disasters, refugee movements, damaged infrastructure or suspicious facilities.
Digital investigations specialist Henk van Ess demonstrated how the tool could add fabricated scenes to genuine locations. Examples included refugees near the United States-Mexico border and a bomb crater near a hospital in Gaza.
Van Ess also reported that an artificial-intelligence detection system failed to recognise some generated material as fake. His tests raised doubts about whether ordinary users could reliably identify manipulated Google Earth images after they were shared outside the platform.
Open-source intelligence researchers depend heavily on satellite images when verifying conflicts, disasters and security incidents.
Google Earth has been widely used to compare buildings, roads, terrain and landmarks with videos or photographs circulating online.
Experts argued that placing an image generator inside the same trusted platform risked blurring the difference between recorded evidence and computer-generated fiction.
They feared that false images could spread faster than researchers could verify or disprove them.
The concern was especially serious during armed conflicts and political crises. A fabricated image of a damaged military site, troop deployment or destroyed civilian facility could mislead the public and influence international debate.
Google Says Images Were Watermarked
Google said the generated images contained digital markers identifying them as artificial-intelligence content.
The company used SynthID technology, which places an invisible watermark inside media created with Google’s artificial-intelligence systems.
Google also clarified that generated images were not added to the main Google Earth database. Other users would not see the altered scenes while browsing the platform.
However, critics argued that these protections were not sufficient.
Users could still take screenshots, crop images or share them without important context. Once posted on social media, the material could appear to be authentic Google Earth evidence.
The platform’s familiar interface could also make manipulated screenshots appear more credible.
Google did not publicly specify which generated images violated its policies. It said some users had shared screenshots that appeared to breach its rules.
The company’s policies prohibit using its artificial-intelligence tools for misinformation, misrepresentation and deliberately misleading activities.
Google said geospatial professionals had also used the feature for legitimate purposes. However, the potential for misuse prompted the company to pause access while stronger safeguards are developed.
Rollback Highlights Wider AI Safety Challenge
The rapid reversal highlights the difficulty technology companies face when introducing generative artificial intelligence into trusted information services.
Image-generation tools can support education, design and planning. They can also create realistic false evidence at high speed and low cost.
The Google Earth controversy was particularly sensitive because the service is widely treated as a factual visual record rather than an entertainment platform.
Researchers said the issue was not limited to whether a digital watermark existed.
The greater concern was how easily generated images could be removed from their original setting and presented as real satellite photographs.
Google’s decision does not necessarily mean the feature has been permanently abandoned. The company said it would work on stronger safeguards before deciding how to proceed.
The “Create Image” button disappeared from Google Earth after the rollback.
The incident has renewed calls for visible labels, stronger restrictions and reliable verification systems for artificial-intelligence-generated media.
It has also raised a larger question for technology companies: whether creative artificial-intelligence tools should be integrated into platforms that people already rely upon as sources of factual evidence.
