Google Shut Down Its Google Earth AI Image Generation Tool After One Day
Google pulled a new AI feature from Google Earth that let users generate satellite imagery via text prompts, citing policy violations and the need for stronger safeguards.
Last verified:
The 24-Hour Lifecycle of Google’s AI Satellite-Image Tool
Google launched and withdrew an artificial-intelligence feature from Google Earth within a single day after independent researchers exposed its inability to prevent the generation of geopolitical deepfakes. According to The Verge, the tool—which used a model called Nano Banana 2—allowed users to modify satellite imagery by submitting text prompts, essentially creating synthetic edits to real-world geospatial data.
Researcher Henk van Ess from Digital Digging demonstrated the tool’s vulnerabilities by deliberately generating false imagery: refugee encampments near the Mexico–United States border and bomb craters adjacent to hospitals in Gaza. Van Ess found that the system declined no requests and offered no alternative suggestions when given sensitive prompts, according to The Verge.
Google’s Removal and Stated Rationale
Google’s initial response to the controversy acknowledged that generated images carried a digital watermark and that the platform filtered requests for harmful content. However, within 24 hours, the company reversed course. In a second statement reported by The Verge, Google acknowledged that “people are sharing screenshots of generated imagery that appear to violate our policies” and announced it was “rolling back this feature in Google Earth while we work on implementing stronger guardrails.”
The company emphasized that generated images remained watermarked and local to individual users—they did not populate the main Google Earth interface for public viewing. This technical isolation, however, proved insufficient to contain the reputational risk, as Van Ess demonstrated that Hive’s automated AI-detection tools could be circumvented by generated video from the feature.
Why This Matters
The rapid withdrawal signals Google’s acknowledgment that watermarks and content-policy disclaimers are inadequate friction against misuse of geospatial synthesis tools. As satellite imagery has become a primary source of real-time intelligence for journalists, humanitarian organizations, and policymakers, synthetic modifications—even if locally generated—present a trust erosion problem for the Google Earth platform itself. Teams relying on Google Earth for verification of remote events now face the compounded problem of distinguishing authentic satellite data from synthetic variants, even when the synthetic content never circulated publicly. The incident suggests that integrating generative AI into geospatial infrastructure requires filtering mechanisms far more restrictive than text-to-image safeguards in consumer applications, or acceptance of the reputational cost of hosting such features at all.
Frequently Asked Questions
What was the Google Earth AI feature that Google shut down?
It was a tool that allowed users to edit satellite imagery in Google Earth by typing text prompts, generating AI-created modifications to real-world imagery using a model called Nano Banana 2.
How long was the feature live?
The feature launched on Thursday and was disabled the following day—approximately 24 hours after release.
Why did Google disable it so quickly?
According to The Verge, independent researchers demonstrated the feature could generate realistic deepfakes of sensitive geopolitical locations (refugee camps, bomb craters) without adequate filtering, and the watermark and content policies proved insufficient.
Did the generated images appear publicly in Google Earth?
No. Google stated that generated images were watermarked as AI-created and did not appear in the main Google Earth experience for other users to see—they remained local to the creator.