Google Earth as a fake machine: why Google stopped its image generator within a day

Google integrated an AI image generator directly into the web version of Google Earth, allowing any user to type a short textual prompt and embed structures, destruction, or events onto a real satellite photograph. Less than 24 hours after launch, it abruptly shut down the feature. The reason was not a technical bug but a safety failure that could have turned the world’s most familiar map into a fertile ground for disinformation.

Nano Banana 2 on a satellite image

The feature relied on the model Nano Banana 2. Once the button was enabled, BBC Verify in Britain and investigative site 404 Media demonstrated how one could create a nuclear core in Iran, a bomb crater in a residential neighborhood, a refugee camp on the US-Mexico border, or the wreckage of a fatal accident on Amsterdam streets. The result was a synthetic image overlaid on genuine geographic data.

Google’s first response explained that every image carries a hidden digital watermark called "SynthID", and users can verify it via the Gemini app or a scan with Google Lens. That argument met sharp criticism. Researchers and disinformation experts in Europe and the United States noted that in combat or political crises, social-media users rarely perform deep fact-checking before sharing content. Moreover, digital watermarks can be bypassed, cut, or disrupted with relative ease. When the network was filled with fake screenshots of disaster zones and battles, Google announced the feature’s removal for a safety review and tighter safeguards.

The experiment highlights the gap between tools intended for entertainment and creation and platforms built on factual ground truth. Image generators such as Midjourney and OpenAI’s DALL-E operate outside a required geographic context, and users understand they are synthetic creations. By contrast, commercial firms like Maxar Technologies and Planet Labs provide high-resolution imagery for governments, research, intelligence and journalism, maintaining strict metadata, provenance chains and precise capture times.

In the past decade, especially since the outbreak of the war in Ukraine and the escalation in the Middle East, open satellite imagery has become the backbone of open-source intelligence (OSINT). Journalists and independent investigators use it daily to verify events, expose war crimes and uncover government propaganda. Google’s combination of an image generator with the platform widely regarded as the primary source for maps and satellites created a crossover that broke the trust walls the public attributes to this information. The swift decision to fold the tool underscores that, in geographic language, map credibility outweighs any technological novelty.