New Delhi: Google has regressed a new AI feature in Google Earth, just hours after its release, after it was feared that it could be used to produce realistic yet fake satellite images.
The feature used Google’s AI image generation model Nano Banana 2, enabling the generation of new images from satellite, aerial and 3D mapping data in Google Earth for all locations. Users were able to zoom in on a location, click on the “Create Image” button, type in a text prompt of what they wanted to view and the image would be created.
The feature was introduced as a creative tool which could enable consumers to visualize places in different time periods, envision construction projects for the future, or sketch in new concepts. The technology might be useful to education, architecture, urban planning and real estate, the company said.
Users might draw a picture of what ancient sites might have been like hundreds of years ago, for instance. Architects and builders could also utilize the feature to envision brand new structures, shopping centers or public spaces prior to their construction.
Researchers, journalists and experts who analyze satellite images, however, soon began to worry about the tool. They stated that AI-generated images associated with actual places could aid in the creation of more realistic fake images of wars, natural disasters, military facilities, among other important events.
These pictures can be disseminated online in a way that gives inaccurate context and can fool viewers into thinking they are satellite images. AI-generated images also posed a risk to the credibility of “real” satellite images and to the ability of journalists and researchers to cross-check information, experts said.
Then, Google said it was taking away the feature for now while it developed more robust safety protocols.
The company said that in an update people have faith in Google Earth as a reliable view of the world. Google said while there were good uses of the feature, some people were posting screenshots of the images Google was generating that were deemed to be in violation of their policies.
Google also explained that the images created by the AI didn’t appear in the primary Google Earth experience and that users wouldn’t have access to them directly in the app. The company stated that each image produced with the feature had an AI label visible, as well as a “SynthID” digital watermark.
SynthID is a digital identifier that Google has invented to help identify AI-generated content. The watermark won’t always be visible to users, but it can serve as a way to let AI detection tools know if the photo was created with Google’s technology.
Google has systems to prevent the creation of harmful content, and it takes misinformation seriously, Google said.
