Bajura Welcome Everyone

Tuesday, August 4, 2026

Google's AI Earth Feature Fails in a Day - Blog Article

Google has made a swift U-turn on an artificial intelligence feature that allowed users to generate artificial images within Google Earth, following a predictable wave of deepfake protests that erupted almost immediately after its launch on July 30, 2026.

The feature, which was accessible to every Google user with no special permissions or waiting list, was designed to leverage the company's Nano Banana 2 image generator. While this AI tool could already produce images from simple text prompts, integrating it directly into Google Earth gave it a dangerous new capability: the ability to use real satellite imagery as a foundation for creating hyper-realistic deepfakes.

This meant users could generate images of buildings, landscapes, and infrastructure with astonishing accuracy — down to the precise shadows, roof textures, and surrounding terrain. What Google presented as a creative tool for students and real estate agents quickly revealed itself as a potential weapon for disinformation .

Within hours of the launch, researchers, journalists, and cybersecurity experts began exposing the feature's dark side, demonstrating how easily it could be used to fabricate evidence of events that never occurred.

The Feature and Its Immediate Fallout

In its initial blog post announcing the launch, Google painted an optimistic picture. "Students could use it to bring history to life," the company suggested, while real estate agents could craft professional property plans with stunning visualizations. Environmental researchers, urban planners, and educators were also mentioned as potential beneficiaries.

But the warnings came fast. Within hours of the feature going live , security researchers and open-source intelligence (OSINT) experts began testing its limits — and the results were alarming.

Henk van Ess , a Dutch researcher specializing in open-source intelligence, was among the first to raise the alarm. "I tried it with refugees at the Mexican border, a nuclear power plant in Iran, a crash in Amsterdam, a hospital with a bomb crater in Gaza," he explained in a series of posts on social media. "Nothing was refused."

Major media outlets quickly followed suit with their own demonstrations. NPR used the tool to fabricate an image of devastating fires sweeping across Iran, as well as a chilling deepfake of Washington, D.C., submerged underwater — images that could easily fuel panic or geopolitical tension if shared without context.

The BBC's verification service, Verify , took the experiment even further. They successfully generated images of the Eiffel Tower collapsing and the Great Pyramid of Giza swallowed by a massive sinkhole . Despite Google's claims that certain guardrails were in place to prevent harmful content, Verify's team managed to circumvent them through basic AI prompt engineering techniques.

What made these demonstrations particularly concerning was the photorealistic quality of the generated images. Unlike earlier AI image generators that produced obvious artifacts or distortions, Nano Banana 2, when combined with Google Earth's real satellite data, created images that were virtually indistinguishable from genuine satellite photography to the untrained eye.

The implications were clear: anyone with a Google account could now create convincing evidence of military buildups, natural disasters, or humanitarian crises in any location on Earth — and share it across social media within seconds.

Google's Response and the Credibility Problem

Faced with mounting criticism and a growing gallery of deepfake examples circulating online, Google acknowledged the failure in a statement posted on X (formerly Twitter):

"We saw geospatial professionals using the feature for various helpful purposes. However, we also saw screenshots of generated images shared by users that appear to break our rules. As a result, we are temporarily disabling this feature in Google Earth while we put stricter guardrails in place."

The company stopped short of committing to never reintroduce the feature. This ambiguity left many observers skeptical, especially given Google's track record of launching AI products that later proved problematic — from Gemini's image generation controversies to earlier AI search missteps.

User reactions were swift and overwhelmingly negative. "It's absolutely impossible that not one person on your development team raised this exact issue," commented one user on the company's announcement post. "You are completely in your bubble," accused another, voicing a sentiment shared by many in the tech community.

Critics pointed out that the risks were entirely foreseeable . The combination of generative AI with authoritative geographic data was always going to attract bad actors. The fact that Google proceeded with the launch anyway raised serious questions about the company's internal review processes...