Transient Illusions: A Brief Glimpse into Google Earth’s AI Deepfake Technology

Google has taken action to disable a recently introduced feature on Google Earth that permitted users to manipulate satellite images through AI-generated text prompts. This tool enabled users to create altered versions of reality by adding elements to images using artificial intelligence. Notably, investigative journalist Henk van Ess experimented with the tool to produce images depicting scenarios like refugees near the Mexican border and a bomb crater near a hospital in Gaza. Google initially responded by stating that images created using the Nano Banana 2 tool in Google Earth would be watermarked and that the company would prevent the creation of images depicting harmful content.

A day later, Google released a new statement explaining the decision to disable the feature. The company acknowledged the trust users place in Google Earth as a reliable source of geographical information. While the feature was utilized by professionals for various legitimate purposes, instances of users sharing manipulated images that violated policies prompted Google to roll back the feature and focus on implementing stricter guidelines. It is important to note that the generated images were watermarked as AI-generated and did not appear in the main Google Earth interface for public viewing.

Van Ess highlighted the potential misuse of the AI-generated images by demonstrating how they could bypass detection tools. He emphasized that the tool allowed him to create misleading content without any restrictions or guidance to reconsider prompt choices.