Google Pulls AI Feature After Two Days: Fake Satellite Images Too Realistic
Set Trending Topics as a preferred source on Google.
Google has shut down a brand-new AI feature in Google Earth after just 48 hours, reports t3n. The reason: BBC journalists managed to bypass the built-in safety measures and create deceptively realistic fake satellite images. The screenshots showed a toppled Eiffel Tower, the Pyramid of Giza sunk into the ground, or a non-existent public park in front of the White House. The generated images were barely distinguishable from real photographs.
Safety Measures Easily Bypassed
The company justified the withdrawal by stating that “people are sharing screenshots of generated images that appear to violate our policies.” While some users were deploying the feature for legitimate purposes — such as geospatial experts — Google says it is nonetheless compelled to develop a version with “stricter safety measures.”
The US tech giant had only launched the AI feature with the image generation model Nano Banana on July 30. The feature was designed to bring ancient sites back to life in three dimensions, visualize construction projects, and add information layers. The generated images were not automatically visible to the public and carried an AI label. Yet BBC journalists quickly found ways to suppress these safeguards — for instance, by simply telling the AI model that the image was real.
High Risk of Fake News
Just two days after launch, Google hit the emergency brake. The speed with which the safety mechanisms could be circumvented highlights the challenge facing tech companies: even well-intentioned features can turn into tools for disinformation if the safeguards are not robust enough.
Data journalist and fake news expert Henk van Ess demonstrated the political explosive potential of the feature to the BBC. It could be used to generate non-existent nuclear reactors in Iran or a refugee camp near the US border. Such fake news carries considerable political impact — especially when dressed in the authoritative guise of Google Earth.
Particularly problematic: AI detection tools used by the BBC themselves were fooled by the fakes in several cases. This makes the generated images a serious threat to information integrity. Google must now go back to the drawing board and make improvements before the feature can make a comeback.
