The Button That Makes Things Up

Google announced a new feature allowing Google Earth users to generate AI visualizations on top of the service's satellite imagery. The "create image" tool, rolled out Thursday, integrates Google's Nano Banana 2 image-generation technology to let users zoom into locations and build images in seconds based on Google Earth satellite, aerial, and 3D-mapping data.[1]

The company unveiled the feature as a way to "visualize history, create real estate plans and more." But disinformation researchers are warning it could enable bad actors to misuse Google Earth, a key resource for fact-checking and verification, to create realistic geospatial fakes. One digital investigations expert wrote: "Google spent 20 years building the reference the world checks against. Today it added a button that makes things up."[2]

That quote is the story. Google Earth is not just a map. It is the map. When journalists need to verify a photo from a conflict zone, when researchers need to confirm whether a building exists at a claimed location, when investigators need to cross-reference a video against known terrain, Google Earth is the reference. It took 20 years to build that trust. It took one update to add a feature that can fabricate what that reference shows.

The key insight from researchers is not that AI-generated images are new. They are not. The problem is context. A fake satellite image on its own might not convince anyone. A fake satellite image layered onto Google Earth, the platform people trust precisely because it shows real imagery, inherits the credibility of the map it was born on. The forgery does not need to be perfect. It needs to be on the right platform.[3]

Google's response is that images created with the feature include SynthID digital watermarks, invisible markers embedded in AI-generated media. Users can ask Google's Gemini chatbot to check if a SynthID is present. The company says it takes misinformation seriously, prevents image creation on harmful topics, and instructs users against misleading activities. AFP's tests on Friday, however, were able to fabricate satellite images that could have serious global consequences.[4]

The gap between "we added watermarks" and "AFP was able to fabricate convincing fake satellite imagery on day one" is where the actual problem lives. Watermarks are a detection mechanism, not a prevention mechanism. They help after the fact, if you know to look for them, if the platform you are viewing the image on supports checking for them, if the person checking knows what SynthID is and has access to Gemini. That is a lot of ifs between a fake image and its detection.

This is the same pattern we have seen all week with AI. Models breaking out of sandboxes. Companies shipping capabilities and adding safety features as an afterthought. The technology is always further along than the safeguards, and the safeguards are always described as robust until someone tests them on day one and finds the gaps. Google Earth was the reference. Now it is also a canvas. Those two functions do not coexist easily, and the trust that took 20 years to build does not survive the discovery that the reference can be edited by anyone with a prompt.