Google Pulls AI Image Feature from Google Earth Amid Misinformation Fears

Google announced Friday it is rolling back a newly introduced AI image-generation feature for Google Earth following user-raised concerns about misinformation. The tool was removed from the company's satellite-imagery program just one day after its introduction, after users generated disturbing and unrealistic images.

The feature, powered by Google's Nano Banana 2 technology, enabled users to create photorealistic images within Google Earth's satellite, aerial, and 3D imagery by inputting text prompts for any global location. This led to the creation of fabricated images depicting events such as bomb blasts in major cities and the appearance of homeless encampments in areas where they did not exist.

Reports indicated that users were able to generate skyscrapers in rural settings, create images of protests outside Google's corporate buildings, and depict war-torn scenarios. Open source intelligence researcher Henk van Ess described generating images of refugees near the Mexican border and a nuclear plant in Iran, stating that the system did not refuse any prompts during his trials.

German investigative reporter Ben Huebel criticized the feature as "absolutely irresponsible and plain bad for investigations," noting its potential to complicate verification work. He warned that even with detection tools, generated images could be manipulated to bypass identification.

Google had initially advertised the generative tool as a means to "visualize history, create real estate plans and more," suggesting it could show users what their city looked like a century ago or what a vacant lot might become as a community garden. However, critics argued that the tool's potential for misuse, particularly in generating realistic images of conflict or social issues, outweighed its intended applications.

In response to mounting criticism, Google acknowledged on X that it takes misinformation seriously and works to prevent image creation on harmful topics, stating that protections are regularly updated. The company also noted that all images generated with Nano Banana 2 in Google Earth were intended to have a SynthID digital watermark to help identify them as AI-generated.

Despite these measures, Google announced the rollback, citing policy violations. While the company did not specify which types of images violated its policies, it stated that it would be working on implementing stronger guardrails to prevent future violations.