Google has integrated its AI image generator, Nano Banana 2, into Google Earth, a rollout that occurred on a Thursday. The feature allows users to generate images by superimposing text-prompted content onto real satellite, aerial, or 3D imagery.
Journalists, human rights advocates, and open source analysts have expressed concerns regarding the potential for abuse and disinformation resulting from this integration. Bill Greer, a geospatial analyst and co-founder of the non-profit satellite service Common Space, stated that the tool is potentially damaging to public trust because satellite imagery has historically been seen as a trustworthy data source.
Testing circumvention of safety guidelines
Test results from BBC Verify showed that small changes to prompts can bypass Google's guidelines intended to prevent the creation of imagery depicting harmful topics. The tool rejected a request to create a 'raised platform to hang traitors' next to the UK Parliament but accepted a less specific prompt to create an image in the same location. Similarly, the tool refused a request to mow the word 'Trump' into the White House lawn, but allowed a community garden to be generated in that spot.
BBC Verify was able to use the feature to create images of a collapsed Eiffel Tower, a sinkhole swallowing the Great Pyramid of Giza, and Russian tanks in Ukraine's capital.
Google is allowing 'invented' imagery to be 'welded to genuine coordinates, drawn on genuine imagery'.
Henk van Ess highlighted risks including the creation of fake images of a non-existent nuclear power plant in Iran, a refugee camp on the US-Mexico border, and a fake hospital in Gaza with a bomb crater. AI-detection researcher Henry Ajdar told BBC Verify that creating fake imagery in conflict zones could be 'destabilising' when information is scarce.
Detection and safeguards
Google stated that they 'take misinformation seriously' and 'prevent image creation on harmful topics and are continually updating our protections'. The company also stated that AI content created by its tools contains invisible watermarks to indicate manipulation.
Users can check if an image is AI-generated using the Lens feature in Google Search or the Gemini chatbot app. However, BBC Verify found it is possible to circumnavigate Google's AI checks and trick the Gemini chatbot into identifying fake Google Earth images as real. External AI detection tools also sometimes failed to identify content generated by the new Google Earth AI feature.
