A nuclear facility in Iran and the bombing in Gaza: Google withdrew from the new tool it launched

In the launch announcement from July 30, Google presented the tool as a means of visualizing history, urban planning and real estate development. Teachers, for example, could turn the remains of Pompeii into a colorful visualization of the city during the Roman Empire. Architects could show how an empty lot might look after the establishment of a residential neighborhood, commercial center or new building. The company announced that the option is available to all users of the web version of Google Earth worldwide.

The cases were not presented as real events, but as an experiment designed to illustrate the risk. Vans claimed that the combination of a real place, precise landmarks and the familiar photography style gives the fake image a particularly believable look. According to him, during the testing process the system did not reject the instructions and did not warn him that these were sensitive requests.

The main concern concerns the use of satellite images for the purpose of verifying information coming from war zones and places to which access is limited. Journalists, intelligence researchers from visible sources and human rights organizations use such photographs to verify reports of attacks, the destruction of buildings, the movement of forces or the construction of facilities. A fake photo based on a real location may be circulated on social media outside of the original context and appear as authentic documentation.

Google initially responded that every image created with the tool included the SynthID digital watermark and that it could be checked using the Gemini app or the visual search tool Google Lens. The company said that it prevents the creation of images on harmful topics and that it takes seriously the spread of false information.

However, digital marking is not necessarily visible to the eye. Screenshots can be cropped, edited or reposted without the information that accompanied the original image. OneS reported that it was able to get an external AI detection tool to misclassify a video created on Google Earth, a finding that illustrates that even automatic detection systems do not provide a definitive answer.

On July 31, Google updated the launch announcement and announced the withdrawal of the tool. The company said that it is aware of the unique trust that the public places in Google Earth and that alongside beneficial professional uses, screenshots of images that seemed to violate its policy were observed. Following this, it was decided to disable the option until stronger protection mechanisms are implemented.

Google did not specify which images were found to be in violation of policy and did not say when the tool might return. The company emphasized that the artificial images were not shown to other users within the main Google Earth experience and did not replace or change the original satellite photos in the database. They were created as separate products by the user who requested to produce them and were marked as images created by artificial intelligence.