Alphabet rolls back AI image generation in Google Earth over policy violations

Alphabet's Google has announced the undoing of an AI-driven image-generation feature it had recently incorporated into Google Earth, following complaints about policy violations. The feature, driven by Google's Nano Banana 2 AI model, permitted users to produce realistic images based on satellite, aerial, and 3D imagery from Google Earth by typing a location prompt.

However, the tool has drawn criticism due to its potential for generating misleading content, enabling users to fabricate scenes over genuine images of monuments and landmarks.

Google acknowledged the valid use of the feature by geospatial professionals for various beneficial purposes, but expressed concern over the sharing of screenshots of generated imagery that seemed to breach their policies. The company has since suspended the capability while it develops more robust safeguards. Unfortunately, the specifics of what types of images contravene Google's policies have not been disclosed.

It should be noted that the images in question are not visible in the standard Google Earth experience and are marked as AI-generated.

The situation echoes a broader trend among tech giants who have integrated AI tools into their products but have subsequently been forced to adjust, restrict, or temporarily disable features due to user violations, misinformation, and policy issues. Earlier this month, Meta decided to discontinue Muse Image, an AI feature that enabled users to generate images using public Instagram accounts, following intense criticism over privacy concerns, including from a Hollywood union.