Google Withdraws New Earth AI Tool After Generating Fake Images of Iranian Nuclear Sites
Google has officially withdrawn its latest artificial intelligence (AI) feature, which allowed users to edit satellite imagery, within only 24 hours of its launch. This move follows demonstrations by users showing that the software could be misused to create misleading visual content.
The tech giant initially announced the integration of the Nano Banana 2 image generation model into Google Earth on Thursday (30/7). The feature was designed to allow users to type any text prompt to virtually reimagine any location in the world overlaid on base layers of satellite, aerial, or 3D imagery.
While Google promoted the feature as a tool to assist with historical visualisation or real estate planning, users quickly tested its limits. OSINT expert and Digital Digging newsletter author, Hank van Ess, revealed that the system failed to filter dangerous topics. Several fake images were successfully created and shared by users, with Van Ess noting that there were no refusals, warnings, or suggestions to use different prompts during his testing.
In response to the spread of screenshots violating safety policies, Google immediately disabled the feature on Friday (3ly/7). The company stated it is working to implement stronger security guardrails. A Google spokesperson explained that although geospatial professionals use this feature for beneficial purposes, misuse by certain parties cannot be ignored. “We are withdrawing this feature from Google Earth while we work to implement stricter protections,” the spokesperson said.
Google emphasised that every image generated by the Nano Banana tool includes a watermark traceable back to their Gemini AI model. Furthermore, user-created images do not appear in the primary Google Earth experience viewed by the general public.
This incident reinforces global concerns regarding the spread of AI-driven misinformation on social media. Previously, AI-generated satellite imagery depicting Iranian drone attacks on US radar systems in Qatar went viral on X (formerly Twitter). As a preventative measure, X has warned content creators of the risk of monetization suspension if they upload AI-generated videos of armed conflicts without clear labelling. Google Earth, which celebrates its 21st anniversary this year, holds over 280 billion images across 110 countries, making it one of the world’s most trusted sources of visual data.
Source: jawawa.id (https://jawawa.id/newsitem/google-withdraws-new-earth-ai-tool-after-generating-fake-images-of-iranian-nuclear-sites-1785603120)
Originally published by Media Indonesia (https://mediaindonesia.com/teknologi/917291/google-tarik-alat-earth-ai-baru-setelah-sajikan-gambar-palsu-nuklir-iran), translated from Indonesian