Google Earth has introduced a new feature that integrates its AI image creation tool, Nano Banana 2, enabling users to generate fabricated satellite and aerial images based on text prompts. This addition has raised significant concerns regarding its potential to be used for misinformation, propaganda, and sowing public distrust.
The functionality allows users to type a prompt and receive a synthetic image, which can then be refined for increased realism. While Google presents the tool as a way to visualize community gardens or recreate historical scenes, early examples suggest a more troubling application. For instance, users have reportedly generated images depicting warzones, riots, and large gatherings of refugees, which, if presented without context, could be used to manipulate public perception.
One notable example involved the creation of an image showing refugees massing at the Mexican border, which appeared convincingly real. Another user reported generating an image of tanks and troops in their hometown. Although the AI-generated images may not always be perfectly realistic, with some instances showing disproportionate elements like tanks as large as buildings, the ability to refine them increases their believability.
Google has stated that it takes misinformation seriously and that all images created with Nano Banana in Google Earth include a SynthID digital watermark, detectable through Gemini or Google Lens. However, concerns remain about the effectiveness of this measure. As noted by Digital Digging's Henk van Ess, detection tools like Hive have struggled to definitively identify AI-generated images, with one analysis returning a 99.2% certainty that a recording of an AI image being created was real, not the image itself. Furthermore, bypassing these detection tools is reportedly straightforward, and many users may not verify the authenticity of images they encounter.
This ease of creating and disseminating potentially false imagery from a trusted platform like Google Earth is a significant worry. While other AI image generators exist, the integration into a service historically relied upon for accurate geographical data lends a veneer of credibility to its outputs, making fabricated content more likely to be believed. Google's response also indicated that it prevents image creation on harmful topics and is continually updating protections, yet users have demonstrated the ability to create warzone scenarios, questioning the efficacy of these safeguards.
The platform's monetization strategy, which limits casual users to a few images per day, may curtail widespread misuse by individuals. However, free trials for Google Cloud users and monthly subscription costs for dedicated users could still provide sufficient access for those with malicious intent. This development positions one of the world's most influential companies as not only capable of disseminating falsehoods but also equipping the public with the means to create their own.
The integration of AI image generation into Google Earth represents a step toward what some describe as an AI dystopia, where the lines between reality and fabrication become increasingly blurred, with potentially serious consequences for public discourse and trust.