Google Earth Withdraws Nano Banana 2 AI Image Feature After Fake Scenes Erode Trust

After launching an AI-powered image generation feature, Google Earth withdrew it within two days after it produced fabricated content. The feature was announced last week for the web version, allowing users to enter prompts at any location on the map and generate new images using satellite, aerial, and 3D imagery as base layers. Google described it as a concept rooted in the real world, envisioning use cases such as community garden renovations and restoration of historical sites. After the feature went live, journalists used a single prompt to generate images of a non-existent nuclear power plant, a border refugee camp, a car crash on an Amsterdam street, and a crater from a field hospital. Verification tests also produced images of a collapsed Eiffel Tower and a sinkhole engulfing the Pyramids of Giza. Google initially responded that the generated images carried SynthID digital watermarks, allowing users to check whether an image was AI-generated via Gemini or Google Lens, and that the system would block harmful content. However, tests found that some dangerous prompts were not intercepted, and that wording modifications could bypass restrictions. When images were screenshotted, cropped, or forwarded, ordinary users found it difficult to verify them. Within less than two days, the feature was withdrawn. Google issued a statement saying it had noticed that people place particular trust in Google Earth's reliable depiction of the world, and that it had seen shared generated images that violated its policies, so it would add stricter safeguards before reconsidering the feature. Since its launch, Google Earth has accumulated about two decades of user trust, with street view imagery covering 10 million miles. Google emphasised that generated images would not appear directly in the Google Earth main view seen by other users, but the spread of misinformation mainly occurred through screenshots and shares on social media platforms. In recent years, Google has integrated generative AI into many products including Search, Gmail, Photos, and Chrome, and the Google Earth integration is part of that push. The company said the feature was originally intended for professional users such as architects and teachers, to place design proposals into real plots or showcase historical cityscapes. But after public release, any user could input arbitrary prompts to generate fictional content. The feature is currently suspended, and Google has not announced a timeline for relaunch.