Google pulled Earth AI image generator after 24-hour backlash

Google's brief experiment with AI-generated satellite imagery

Google removed a generative AI feature from Google Earth within 24 hours of its July 30 launch after users created fabricated satellite images of nuclear facilities, bomb craters, and disaster scenes—all blended seamlessly into real geographic coordinates.

The feature allowed anyone to zoom to actual locations and prompt AI to generate photorealistic scenes overlaid on trusted satellite imagery. Within hours, fake images appeared showing a nuclear plant in Iran, a bomb crater beside a Gaza hospital, and floods at the U.S. Capitol, according to reporting first published by Rest of World.

Google pulled the feature following criticism from investigative experts and media outlets, promising to return it with "stronger guardrails." But the episode highlighted a troubling pattern: AI fabrication tools launch without adequate safeguards, cause immediate harm during sensitive moments, and only then get restricted.

Why it matters

Google Earth has served as foundational verification infrastructure for two decades. Open-source investigators rely on it for conflict documentation in Syria, Yemen, and Sudan. Integrating image generation directly into this trusted platform—especially during an active conflict between Israel, the U.S., and Iran—compromised the tool that journalists and analysts depend on to separate fact from fiction. The damage to trust persists even after the feature's removal, as fabricated images continue circulating and authentic satellite evidence now carries undeserved doubt.

The broader pattern of AI harm

This wasn't an isolated incident. OpenAI shut down its Sora video app in March 2025 after months of abusive content. Google's Veo 3 video generator, launched in May 2025, quickly produced fabricated footage of missile strikes on Tel Aviv during the Iran conflict, with the company's watermark still visible.

Watermarks and detection tools remain inadequate. When Witness tested detection capabilities across 13 AI providers, seven had no public detection tool, and nearly all that did could be fooled. Meanwhile, fabricated footage of a blast at Iran's Evin prison spread through major outlets including Sky News and the BBC before experts could verify it as AI-generated days later.

The U.S. government simultaneously restricted journalists' access to commercial satellite imagery of the conflict region, citing wartime security—narrowing the supply of authentic information as counterfeits proliferated.

What effective safeguards look like

Some manufacturers are building verification into hardware. Google's Pixel 10 became the first smartphone to integrate the C2PA provenance standard directly into its camera, cryptographically signing photos at capture. Apple's iPhone 18 Pro models include features to produce signed reference images and identify AI-generated or edited content.

Deepfake expert Hany Farid noted that when major companies address this problem at the hardware level, it acknowledges that loss of trust in visual evidence has become a mainstream crisis requiring mainstream solutions.

Regulation lags behind. The EU AI Act's transparency obligations began in August 2026, but precise rules remain voluntary rather than enforceable. California's AI transparency law won't reach platforms until at least 2027.

The liar's dividend in practice

The concept of the "liar's dividend"—where the existence of convincing fakes allows bad actors to dismiss authentic evidence—has moved from theory to operational reality in modern conflicts. Fabrications don't need to survive scrutiny; they only need to exist long enough for doubt to take hold before the truth arrives.

Google stated the feature will return once stronger guardrails are ready. But as Mahsa Alimardani, associate director of the technology threats and opportunities program at Witness, argued in the original Rest of World analysis: no benefit of integrating generative AI into Google Earth justifies the harm and doubt it creates.

The fabricated satellite frames from this conflict continue circulating in new claims, while authentic documentation now carries an asterisk it doesn't deserve.

Details of the Google Earth AI feature and its removal were first reported by Mahsa Alimardani at Rest of World.