Full Breakdown
Google Earth’s AI Image Generator Pulled After One Day Amid Disinformation Fears
8/5/2026, 4:24:37 AM
Google Earth Introduces, Then Withdraws AI Image Generator
On July 30, Google added an artificial-intelligence tool called “Nano Banana 2” to Google Earth, allowing users to type a text prompt and generate a new image over any satellite, aerial or 3D view. Within hours, open-source intelligence researchers and journalists demonstrated that the feature could produce realistic-looking scenes such as refugees at the Mexico-U.S. border, a nuclear plant in Iran, a bomb crater in Gaza, and a collapsed Eiffel Tower.
By July 31, Google announced that it was rolling back the feature while it worked on stronger safeguards. The company said the generated images never appeared in the default Google Earth view for other users and were watermarked as AI-generated, but critics warned that screenshots could be shared without the watermark.
Background & Context
For two decades Google Earth has been a trusted reference for journalists, governments and investigators verifying events in inaccessible locations. Satellite imagery has been used to expose China’s Uyghur detention camps, document destruction in Gaza, and reveal human-rights abuses in Sudan. The addition of a generative-AI overlay threatened to blur the line between verified satellite data and fabricated visualizations.
Criticism & Opposition
Evan Hill, a visual-forensics reporter for *The Washington Post*, warned that the tool’s ability to create “fictionally realistic” scenes could make verification “effectively limitless,” especially when screenshots are shared without context.
Other experts highlighted that the feature could become a “disinformation nightmare” because satellite imagery carries an inherent perception of authority, making fabricated scenes appear more credible than ordinary AI-generated pictures.
Conflicting Reports & Gaps
Google asserts that every AI-generated image includes a SynthID digital watermark that can be checked via the Gemini app or Google Lens. The BBC reported that its tests found the watermark effective for identifying AI images. Conversely, an X community note claimed that the watermark is applied only to portions of the generated image, allowing users to circumvent verification tools. No independent study has yet measured how often the watermark is removed or obscured in online sharing.
Why It Matters
The episode illustrates a new vector for visual disinformation: realistic, location-specific deepfakes that can be mistaken for authentic satellite evidence. Because mapping platforms are routinely used to corroborate reports of wars, natural disasters and human-rights violations, the ability to fabricate such images threatens journalistic verification, humanitarian monitoring and public trust in geospatial data. The incident also underscores the absence of industry-wide standards for safeguarding the credibility of AI-enhanced mapping services.
What’s Next
Google has not announced a timeline for re-introducing the feature. The company indicated it will develop “stronger guardrails,” potentially including tighter prompt restrictions on conflicts, disasters and sensitive infrastructure, and more durable methods for identifying generated images after they leave the platform.
Verbatim Quotes
- “We have seen professionals in the geospatial sector use this feature for a range of useful purposes,” — Google Earth Google (Google Earth, Ilsole24ore)
