New Google Earth AI tool could fuel misinformation, experts say

Google has pulled its recently launched AI feature for Google Earth after reports highlighted significant risks regarding the creation of misleading content. The tool allowed users to generate artificial scenarios atop real satellite and aerial imagery using text prompts. While the intent was to provide new creative capabilities, tests by BBC Verify demonstrated that the system could easily produce fake evidence of destruction, such as a collapsed Eiffel Tower or damage to historical sites.

The feature existed for less than 48 hours before Google announced a pause to implement stronger safety measures. Experts noted that because these AI-generated visuals were anchored to genuine geographic coordinates, they carried an unearned sense of credibility. This reliance on the perceived accuracy of Google Earth meant that manipulated images could be presented as verified data by bad actors.

Researchers identified several ways to bypass the company's internal guidelines. By adjusting prompts, users could trick the system into generating restricted or harmful content that the safety filters were meant to block. Even though Google includes invisible watermarks on AI-generated output, researchers found these could be ignored or misinterpreted by other detection tools.

This incident highlights a broader tension between the rapid deployment of AI tools and the preservation of objective information. As people look to satellite imagery for reliable insights into global events, the erosion of trust in these sources poses a risk to journalists and the public. Google acknowledged that users expect a reliable view of the world from its platform and stated that it is prioritizing the integrity of its mapping services as it re-evaluates the tool.