Image Courtesy: Google
Google has rolled back the AI image generation feature it recently introduced in Google Earth, less than 24 hours after announcing the integration. The decision follows concerns that the tool could be used to create convincing but entirely fabricated satellite imagery, raising fresh questions about AI-generated misinformation.
The feature, powered by Google’s Nano Banana image generation model, allowed users to create synthetic imagery directly within Google Earth. While Google said the capability had legitimate applications for geospatial professionals, it acknowledged that some users quickly began sharing AI-generated images that appeared to violate the company’s policies. The reversal came after criticism over the tool’s potential for misuse.
Concerns intensified after 404 Media demonstrated how easily the feature could generate misleading satellite-style images. In one example, journalist Joseph Cox created a fake blast crater in Los Angeles. Another image depicted protesters gathered outside Google’s headquarters in Mountain View, California, despite no such event occurring.
Before removing the feature, Google defended the system by pointing to its built-in safeguards. The company said every AI-generated image included a SynthID digital watermark and could be identified using Google Lens or the Gemini app. It also said the tool blocked image generation for harmful topics and that its safety protections were continuously being updated.
Critics argued those measures were insufficient. Cox responded that relying on users to verify images misunderstands how misinformation spreads online, noting that most people rarely stop to authenticate images before sharing them.
In a statement posted on X, Google reaffirmed that Nano Banana images are digitally watermarked but said it would introduce stronger guardrails before bringing the feature back to Google Earth. The company did not provide a timeline for the tool’s return.
The rapid rollback highlights the growing challenge technology companies face as increasingly realistic AI image generators become widely available. While digital watermarking may help identify synthetic content after the fact, preventing convincing fake images from spreading remains a far more difficult problem.
