Google Earth's AI deepfake tool only lasted one day
News Source
•2026-07-31T15:13:29-04:00
🔍 The Backstory
Google Earth launched a new feature on Thursday that allowed users to edit satellite images with text prompts using AI, powered by an image generation model called Nano Banana 2. The tool essentially let anyone create AI deepfakes of real-world locations — adding, removing, or modifying buildings, landscapes, and landmarks through simple text descriptions. Google touted it as a creative way to reimagine the world.
Within hours, researchers and journalists demonstrated the tool's potential for abuse. Henk van Ess of Digital Digging intentionally generated images adding refugees near the Mexican border and a bomb crater by a hospital in Gaza, proving how easily the feature could be weaponised for disinformation. Google initially responded that the images were watermarked, but within 24 hours the feature was shut down entirely — an extraordinarily short lifespan for a major product launch from one of the world's largest technology companies.
🎯 Why It Matters
This incident is a dramatic example of the tension between AI innovation and responsible deployment. Google Earth's AI deepfake tool was conceptually impressive but catastrophically poorly thought out from a safety perspective. The fact that it took external researchers — not Google's own safety teams — to demonstrate the obvious misuse potential within hours of launch suggests fundamental failures in the company's AI review processes.
The rapid shutdown also demonstrates the power of public scrutiny and the risks companies face when they prioritise shipping features over safety. For Google, which has positioned itself as a responsible AI leader, this is an embarrassing episode that will likely have internal repercussions and could slow future product launches.
Google Earth launched a new feature on Thursday that allowed users to edit satellite images with text prompts using AI, powered by an image generation model called Nano Banana 2. The tool essentially let anyone create AI deepfakes of real-world locations — adding, removing, or modifying buildings, landscapes, and landmarks through simple text descriptions. Google touted it as a creative way to reimagine the world.
Within hours, researchers and journalists demonstrated the tool's potential for abuse. Henk van Ess of Digital Digging intentionally generated images adding refugees near the Mexican border and a bomb crater by a hospital in Gaza, proving how easily the feature could be weaponised for disinformation. Google initially responded that the images were watermarked, but within 24 hours the feature was shut down entirely — an extraordinarily short lifespan for a major product launch from one of the world's largest technology companies.
This incident is a dramatic example of the tension between AI innovation and responsible deployment. Google Earth's AI deepfake tool was conceptually impressive but catastrophically poorly thought out from a safety perspective. The fact that it took external researchers — not Google's own safety teams — to demonstrate the obvious misuse potential within hours of launch suggests fundamental failures in the company's AI review processes.
The rapid shutdown also demonstrates the power of public scrutiny and the risks companies face when they prioritise shipping features over safety. For Google, which has positioned itself as a responsible AI leader, this is an embarrassing episode that will likely have internal repercussions and could slow future product launches.