Why Google Earth’s AI image generator is a problem
▼ Summary
– Google Earth’s new AI feature, Nano Banana, lets users generate altered satellite and 3D images from text prompts, such as fake scenes of refugees or bomb craters.
– Google claims every AI-generated image includes a SynthID digital watermark and can be verified via the Gemini app or Lens in Search.
– Researcher Henk van Ess suggests cross-checking suspicious images with other satellite platforms like Sentinel-2 or Landsat and verifying orbital metadata.
– The feature launched globally on the web version of Google Earth, pitched as a tool for visualizing historical sites and real estate projects.
– Despite the watermark, some AI-altered images are believable enough to spread misinformation, and Digital Digging fooled Hive’s AI detector with an AI-altered video.
A simple text prompt can now conjure altered realities from Google Earth’s satellite and aerial imagery. Digital Digging researcher Henk van Ess demonstrated this by generating convincing fakes of “refugees near the Mexican border” and a bomb crater outside a hospital in Gaza, using the platform’s new AI tool.
Google has pushed back on the concerns, stating, “We take misinformation seriously. Every image created with Nano Banana in Google Earth includes the SynthID digital watermark, so if someone is unsure about an image, they can ask the Gemini app or use Lens in Search to see if the image was AI-generated.” Van Ess, however, advises a broader verification strategy. Beyond running suspicious files through Google’s “@verifyai” tag in Gemini, he recommends cross-checking the content against independent satellite platforms like Sentinel-2 or Landsat, and scrutinizing the metadata for orbital details, such as the capture time and the specific satellite involved.
The feature rolled out globally on the web version of Google Earth earlier this week. Google’s promotional material frames it as a creative aid for visualizing historical landmarks or prospective real estate developments. Some outputs are obviously fabricated, for instance, a Sphinx fused with the Statue of Liberty. Yet the underlying risk is real. The Nano Banana 2 engine can produce imagery that is plausible enough to slip past casual scrutiny and fuel viral misinformation.
The safeguards are not foolproof. While the SynthID watermark is baked into every generated image, removing it is difficult but not impossible, and casual viewers may simply never look for it. Digital Digging already proved the system’s limits by tricking Hive’s AI detector with an AI-altered video sourced from Google Earth, underscoring how easily these tools can outpace the very defenses meant to contain them.
(Source: The Verge)
