Google rolled back its AI image generation feature for its satellite mapping service, Google Earth, just one day after its introduction. The decision follows concerns that the tool could be misused to spread misinformation by overlaying fake scenes onto real terrain and buildings.
On August 1, Alphabet's subsidiary Google announced the integration of its AI image generation tool, Nano Banana 2, into Google Earth. However, the following day, it announced the withdrawal of this feature.
The tool allowed users to zoom in on specific areas using Google Earth's satellite, aerial, and 3D map data and generate AI images in seconds. Brian Horowitz, the product manager for Google Earth, highlighted potential uses for the feature, including visualizing historical scenes, drafting real estate plans, and visualizing construction plans before breaking ground.
However, open-source intelligence (OSINT) experts and misinformation researchers strongly opposed the feature, arguing that it could complicate fact-checking by superimposing plausible fake scenes onto actual images of terrain and buildings. Google Earth has been a critical resource for journalists and researchers verifying damage in conflict zones, restricted areas, military facilities, and disaster sites.
The Financial Times reported that fake images generated by the feature, including bomb craters in Gaza, Iranian nuclear facilities, and car accidents in Amsterdam, were posted on social media platform X. AFP also reported that its tests produced fake satellite images reminiscent of sensitive international issues, such as explosions in Paris, Iranian nuclear sites, Russian bomb craters, and Islamic State training camps in Syria.
As the controversy grew, Horowitz announced the rollback less than 24 hours after the feature's launch. He stated, "The trust people have in Google Earth is unparalleled," adding that while there were instances of geospatial experts using the feature for beneficial purposes, there were also cases of generated images being shared that appeared to violate policies. He noted that the feature would be reverted while stronger safety measures are developed.
AI-edited images were not visible on the default Google Earth interface for other users and were marked with an AI-generated watermark. However, experts pointed out that these measures alone were insufficient to prevent misuse.
Reuters reported that Meta also introduced an AI image generation feature called Muse Image on Instagram earlier this month, only to suspend it due to privacy concerns.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.