Google Earth’s AI makeover lasted a day before deepfake fears won

Google rapidly withdrew a Google Earth image-generation tool after critics showed how easily satellite-style visuals could be fabricated. The company says it will return only after stronger safeguards are in place.
Google Earth’s AI makeover lasted a day before deepfake fears won

Google Earth’s AI makeover lasted a day before deepfake fears won
Google wanted to make Earth imagery more imaginative. Within a day, critics had turned that same promise into a warning about how quickly trusted maps could become engines for believable fiction.

The feature launched Thursday, bringing Nano Banana 2 image generation into Google Earth and inviting users to reimagine places—from historic city views to a proposed community basketball court. Demis Hassabis amplified Google AI’s launch message on X, framing the tool as a way to combine satellite imagery with creative visualization.

But the rollout immediately collided with the special authority of Google Earth. Critics argued that a prompt-based tool capable of placing new objects over real locations could manufacture persuasive “evidence” for social media, journalists and researchers. Henk van Ess tested the system by generating scenes including refugees near the Mexican border and a bomb crater near a Gaza hospital; he said, “Nothing was refused, nothing was softened, and nothing suggested I try a different prompt.”

Google initially stressed that generated images carried a digital watermark and that harmful requests were restricted. Yet van Ess said he could make a Google Earth-generated video evade Hive’s AI detector, illustrating the gap between a label attached to content and the way altered imagery may travel online.

By Friday, Google had rolled the feature back. The company said geospatial professionals had found useful applications, but acknowledged that screenshots “appear to violate our policies” and said it was working on “stronger guardrails.” Google also maintained that generated images were not visible in the main Earth experience for other users and were marked as AI-made.

The dispute is bigger than one short-lived product: synthetic imagery is already easy to produce across the web. But critics’ point was sharper—when the altered picture carries the visual language of Google Earth, the reputational cost of getting it wrong rises fast. A separate account of the episode similarly cast the pause as temporary, with Google expected to revisit the tool with added verification and safeguards.

Continue reading https://foxvector.com/stories/019fe443-89b4-222e-7267-02b625e02346

Write a comment