Google abruptly removed an image generation tool from Google Earth on July 31, 2026, just one day after its release. The feature allowed users to create AI-powered satellite visuals with simple text prompts, sparking immediate alarm among open-source researchers over the rapid proliferation of dangerous deepfakes.
The rollout began on Thursday, July 30, 2026, when Google introduced Nano Banana image generation capabilities to its Google Earth platform. The feature was designed to let users blend custom visuals with real-world satellite, aerial, and 3D data. Product manager Bryan Horowitz explained in the initial announcement that users could simply zoom in to a place in Google Earth on web, tap ‘create image,’ and type whatever you want to see.
Company marketing materials suggested benign educational and creative uses, such as visualizing historic Roman Empire street scenes over modern-day Pompeii ruins or picturing a community garden in an empty neighborhood lot. But within hours of launch, open-source investigators and digital forensic experts discovered the tool possessed zero guardrails against manufacturing plausible geopolitical crises.
Testing the Limits: Faking Disasters and Military Conflict
Open-source researchers quickly tested the boundaries of the new tool and found it alarmingly permissive. Henk van Ess, an open-source researcher who first exposed the platform’s vulnerabilities, tested several high-stakes scenarios via text message with NPR.
“I tried refugees at the Mexican border, a nuclear plant in Iran, a crash in Amsterdam, a hospital with a bomb crater in Gaza. Nothing was refused.”
Henk van Ess, open-source researcher
Journalists and investigators demonstrated the system’s volatility by easily generating realistic-looking imagery of Iran’s Kharg Island engulfed in fires and a severely flooded U.S. Capitol complex. Neither event occurred, yet both would represent major breaking news if authentic.
The ease of generation stunned professionals who monitor global conflict. Evan Hill, a visual forensics investigator at the Washington Post, questioned on X whether the company had conducted any internal safety reviews before deployment, noting that the opportunities for abuse and disinfo are literally boundless.
The Stakes for Verification and Global Trust
For years, authentic satellite imagery has served as a cornerstone for verifying breaking news, military strikes, and humanitarian atrocities in closed or dangerous regions. Because capturing high-resolution overhead images historically required complex orbital hardware controlled by governments or major institutions, space-based photography remained a trusted baseline.
Jake Godin, a senior researcher with the visual forensics group Bellingcat, emphasized that satellite imagery has been kind of a safe bet when it comes to verifying an event because it’s hard to fake.
Fabricated overhead imagery is not entirely unprecedented. Earlier in the U.S.-Israel war with Iran, fraudulent images circulated in Iranian media showing a damaged American military base in Bahrain by manipulating existing Google Earth visuals. However, packaging text-to-image AI directly into a ubiquitous mapping platform threatens to scale that deception dramatically.
It’s just streamlining the process, which I think is going to make it proliferate more, Godin warned, noting that misinformation moves faster than corrections and that widespread fakes will allow bad actors to dismiss real photographic evidence as fabricated.
Google’s Rollback and Safety Safeguards
Faced with mounting criticism, Google reversed course on Friday, July 31, 2026. In a statement posted to X, the company acknowledged the utility for geospatial professionals but pulled the plug due to policy-violating screenshots.
“We’ve seen geospatial professionals using this feature for a range of useful purposes. However we’ve also seen people sharing screenshots of generated imagery that appear to violate our policies. So we’re rolling back this feature in Google Earth while we work on implementing stronger guardrails.”
Google, via statement on X
The tech giant stressed that it takes misinformation seriously and embeds its SynthID watermark into images to ensure they register as machine-generated when processed by tools like Gemini. The company also maintained that in addition, we prevent image creation on harmful topics and are continually updating our protections.
Despite these assurances, open-source investigators remain skeptical that technical watermarks or rudimentary keyword blocks can secure mapping software against sophisticated abuse, leaving the future availability of the tool uncertain as developers head back to the drawing board.
