Google Earth’s AI deepfake tool only lasted one day
Google has rolled back its newly launched satellite image editing feature in Google Earth after users quickly leveraged AI text prompts to generate deceptive deepfakes.
Stock photo for illustration only, not from the actual event
- Google Earth shut down its AI image editing feature just one day after launch
- The tool allowed users to modify satellite imagery using text prompts via Nano Banana 2
- Digital researchers successfully tested it by generating fake conflicts and refugee images
- Google stated it is rolling back the feature to implement stronger guardrails and safety policies
Granting everyday users the ability to effortlessly modify satellite imagery using AI text prompts proved to be a critical misstep. Google has officially pulled the plug on a newly introduced Google Earth feature, which debuted on Thursday, allowing individuals to alter real-world satellite views through generative artificial intelligence.
Essentially, the tool enabled users to manufacture realistic AI deepfakes of physical locations using simple text commands. For instance, Henk van Ess from Digital Digging intentionally generated fabricated imagery depicting refugees near the Mexican border and a simulated bomb crater situated next to a hospital in Gaza.
Stock photo for illustration only, not from the actual event
Initially, on Thursday, Google defended the rollout by pointing out that images generated via Nano Banana 2 inside Google Earth featured a digital watermark and that the corporation actively prevented image creation concerning harmful topics.
However, just one day later, Google issued a revised statement noting that they recognize people uniquely trust Google Earth for a reliable view of the world. While geospatial professionals utilized the feature for useful applications, the company observed users sharing screenshots of generated imagery that appeared to violate its policies. Consequently, Google decided to roll back the feature while working on implementing stronger guardrails, noting that generated images never appeared in the main Google Earth experience and always carried AI watermarks.
The rapid withdrawal of this AI mapping feature highlights the immense pressure and responsibility tech giants face when deploying generative AI tools into mainstream public platforms. When powerful image-generation models are merged with global geographical data, the potential for misinformation and geopolitical manipulation multiplies exponentially, forcing companies to prioritize strict verification guardrails over rapid feature deployment.
Furthermore, van Ess demonstrated that he successfully bypassed Hive's AI detection tool using a video generated directly within Google Earth, proving that synthetic media can continue to spread despite embedded watermarks. In his testing remarks, van Ess noted that nothing was refused, nothing was softened, and nothing suggested trying an alternative prompt.
Source: The Verge
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment