Google has withdrawn an artificial intelligence tool that let users paint fabricated scenes on top of real satellite imagery in Google Earth, reversing course less than 48 hours after launch amid warnings that the feature could supercharge misinformation.
The feature, powered by Google's "Nano Banana 2" image generator, was introduced on Thursday, July 30, 2026, and allowed anyone to type a text prompt and generate false scenarios layered over genuine satellite, aerial, and 3D imagery. For the latest developments on AI tools shaping the industry, follow our breaking AI news coverage at AI Buzz Wire.
By Friday, Google confirmed it was pausing the capability "while we work on implementing stronger guardrails," citing screenshots of generated imagery that appeared to violate its policies. The abrupt reversal came directly after a BBC Verify investigation in which researchers demonstrated how easily the tool could produce convincing, location-accurate fakes.
How the Tool Worked
Nano Banana 2 integrated directly into Google Earth, the widely used mapping platform that governments, journalists, and the public rely on for an accurate view of the planet. The AI feature generated imagery on demand and placed it over a base layer of real coordinates. Crucially, the resulting images could be downloaded and shared on social media and other platforms with no obvious indication they were synthetic.
That combination — invented visuals fused to genuine geographic data — is what alarmed researchers most.
Alarming Demonstrations
During testing, BBC Verify was able to generate imagery of a collapsed Eiffel Tower, a sinkhole swallowing the Great Pyramid of Giza, and Russian tanks positioned in Ukraine's capital, Kyiv.
AI and misinformation researcher Henk van Ess went further before the tool was suspended, producing fabricated scenes of a non-existent nuclear power plant in Iran, a refugee camp on the US-Mexico border, and a fake hospital in Gaza with a bomb crater beside it.
"The forgery does not have to look convincing on its own. It inherits the credibility of the map it was born on," van Ess said, describing how Google had allowed "invented" imagery to be "welded to genuine coordinates, drawn on genuine imagery."
Watermarks That Didn't Hold Up
Google has emphasized that AI-generated content from its tools carries invisible watermarks and that users can verify suspected fakes through the Gemini chatbot or Google Lens.
However, BBC Verify's testing found it was possible to circumvent those checks and trick Gemini into confirming fabricated Google Earth images as real. External AI detection tools also failed to flag some of the synthetic imagery. The findings undercut the reliability of the safeguards Google has positioned as a defense against AI-driven misinformation.
AI-detection researcher Henry Ajder told BBC Verify that fabricated imagery in conflict zones could prove especially "destabilising" during fast-moving events where reliable information is scarce and the stakes are highest.
A Pattern of Overcorrection
The episode adds to a growing list of high-profile AI launches that companies have had to walk back. Google itself faced scrutiny over inaccurate image generation in its Gemini model in 2024, and the rapid deployment-then-retraction cycle has become a recurring theme as technology firms race to ship generative features.
In this case, the unique danger stemmed from the platform itself. As Google acknowledged in its statement, "We know that people uniquely trust Google Earth for a reliable view of the world" — making any erosion of that trust especially costly.
The rollback also comes at a sensitive regulatory moment. The European Union's AI Act transparency requirements, which mandate clear labeling of AI-generated content, took effect on August 2, 2026, just days after the Google Earth incident. Google had already signed the EU's code of practice on AI-generated content labeling earlier in July.
What Happens Next
Google has not said when, or whether, the feature might return. The company indicated it would continue refining guardrails, but researchers have argued that the fundamental design — layering AI-generated content over trusted geographic data — may be too dangerous to ship in a downloadable, shareable form without far stronger provenance controls.
For an industry still grappling with how to balance rapid AI deployment against real-world harm, the 48-hour lifespan of Google Earth's AI image tool stands as a cautionary tale about the gap between what AI can create and what it should.
Stay Ahead of AI
AI is moving fast, and the policies shaping it move even faster. Bookmark AI Buzz Wire for daily, fact-checked coverage of the launches, reversals, and regulations defining the future of artificial intelligence.
Read more AI news →
