The shortest path between truth and propaganda just got a lot shorter.
The Summary
- Google added Nano Banana AI image generation to Google Earth, letting users create fake satellite imagery from text prompts
- Digital Digging's Henk van Ess demonstrated the risk by generating images of "refugees near the Mexican border" and fake bomb craters near Gaza hospitals
- Google's defense: every generated image includes SynthID watermarking, detectable through Gemini or Lens
- The trust infrastructure for satellite imagery — evidence in war crimes tribunals, border policy debates, environmental monitoring — now requires technical literacy to verify
The Signal
Google just handed the world's most trusted mapping platform a feature that makes misinformation trivially easy to create. Nano Banana turns text prompts into photorealistic fake satellite images using Google Earth's existing visual database. Type "refugees near the Mexican border" and get an image that looks like it came from an intelligence briefing. Type "bomb crater near Gaza hospital" and manufacture evidence of a war crime.
The examples from Henk van Ess at Digital Digging aren't edge cases. They're the obvious use cases. Satellite imagery carries institutional weight. Newsrooms use it. Human rights organizations cite it. Governments make policy decisions from it. That credibility took decades to build. Google's letting users burn it for fun.
"Every image created with Nano Banana includes the SynthID digital watermark."
Google's response assumes three things that don't hold:
- Users know to verify images before sharing them
- Platforms and people have adopted SynthID verification tools
- Watermarks survive screenshot, re-upload, and cross-platform sharing
None of those are true at scale. Most people don't verify. Most platforms don't check. Watermarks are fragile. The ratio of image creation to image verification is 1000:1, minimum. By the time a fact-checker catches a fake satellite image, it's already shaped a million impressions.
The real problem isn't the technology. Generative AI was going to reach satellite imagery eventually. The problem is Google's distribution strategy. They're not testing this with researchers or journalists who understand verification chains. They added it directly to Google Earth, a consumer product used by hundreds of millions of people, with a whimsical name that undersells the risk.
The Implication
We're entering a world where satellite imagery needs the same skepticism we apply to text. Provenance becomes everything. If you work in media, policy, or humanitarian response, update your verification protocols now. Assume any satellite image without a direct institutional source could be synthetic. Build relationships with organizations that control primary satellite feeds.
For the rest of us: the "@verifyai" tag Google mentions is your new reflex. Before you share an overhead image that confirms your priors about a border, a war zone, or an environmental disaster, verify it. Because the next thousand fake images won't come from Digital Digging. They'll come from people who don't know what they're doing or don't care.