Google Scraps AI Satellite Images Within 24 Hours: Why 'Deepfakes in the Sky' Alarmed Researchers

Google removed its AI-powered image generation feature from Google Earth less than 24 hours after launch, following backlash from researchers and open-source intelligence experts who warned it could spread misinformation by fabricating satellite images of disasters and military sites. The feature, demonstrated by product manager Bryan Horowitz, allowed users to generate AI-created satellite imagery from text prompts, but critics like Henk van Ess and Evan Hill said it was insufficiently tested and could undermine trust in satellite imagery as evidence. Google confirmed the rollback and said it would introduce stronger safeguards.

Google Scraps AI Satellite Images Within 24 Hours: Why 'Deepfakes in the Sky' Alarmed Researchers The AI could generate disasters that never happened and that was enough to force a rapid rollback Google has shut down its Google Earth image generation feature less than 24 hours after introducing it. This happened after facing fierce backlash from journalists, researchers, and open-source intelligence experts who warned that it could accelerate the spread of misinformation. The feature allowed users to generate AI-created satellite images by zooming into any location in Google Earth and entering a text prompt. Google described the feature as a creative way to reimagine places around the world. However, critics warned it could weaken trust in Google Earth, which has become an important tool for journalists investigating conflicts, natural disasters, and other major events. The tech giant later confirmed that it was rolling back the feature while introducing stronger safeguards and acknowledging that some AI-generated images appeared to fall outside its policies. A Creative Tool That Raised Red Flags Recently, Google launched the feature to allow users to simply zoom into a location, tap 'Create image', and describe the scene they wanted to generate. Product manager Bryan Horowitz demonstrated the process during the announcement. That flexibility immediately alarmed researchers. They quickly found the tool could generate fabricated satellite images depicting disasters, military sites, and humanitarian crises with few apparent restrictions. Open-source investigator Henk van Ess said that even highly sensitive prompts were approved. '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,' he said. For journalists and investigators who rely on satellite imagery to verify real-world events, the findings were deeply concerning. Visual forensics investigator Evan Hill also questioned whether Google had adequately tested the feature before launch, writing on X: 'Very curious how or if this idea was red teamed internally because the opportunities for abuse and disinfo are literally boundless.' Why Satellite Images Are Different Unlike typical AI-generated https://www.ibtimes.co.uk/ai-generated-health-experts-concern-1810987 images, satellite imagery is widely regarded as one of the most trusted forms of visual evidence. Journalists and investigators often cross-check satellite imagery with videos and social media posts to verify destroyed buildings, military movements, and the impact of natural disasters. Although Google Earth's imagery isn't always the most up-to-date, its worldwide coverage has made it a valuable resource for journalists and organisations conducting visual investigations. Jake Godin, a senior researcher at Bellingcat, summed up why that trust matters. 'Satellite imagery has been kind of a safe bet when it comes to verifying an event because it's hard to fake,' he said. 'Deepfakes in the Sky' Fabricated satellite images are nothing new. During the Iran-Israel conflict, AI-edited images falsely showing damage to a US military base in Bahrain spread online before real satellite photos emerged. Those fakes, however, were created by editing genuine satellite images using AI tools. Google's new feature removed that extra step by allowing users to generate convincing satellite-style imagery from a text prompt. Godin warned that the simplified process could speed up the spread of false content: 'It's just streamlining the process, which I think is going to make it proliferate more.' Experts also warn that as AI-generated satellite imagery becomes more widespread, real evidence could be dismissed as AI-generated. As Godin noted, misinformation often spreads much faster than the truth. Google Hits Reverse Google swiftly pulled the feature after widespread criticism. 'We've seen geospatial professionals using this feature for a range of useful purposes,' the company said on X. '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 previously said that its AI-generated images include its SynthID watermark and can be detected by products such as Gemini https://www.ibtimes.co.uk/apple-wwdc-revamped-siri-ai-system-upgrades-1801584 , which are subject to safeguards designed to block harmful content. However, those assurances failed to convince many researchers. Journalist Shayan Sardarizadeh mocked the rollout, sarcastically saying that there was 'no way' a tool like this could be used to spread misinformation through one of the internet's most trusted visual platforms. Meanwhile, Henk van Ess warned that adding AI-generated images to Google Earth could undermine its reputation as a trusted resource that has long been used by investigators to verify social media claims. The rapid rollback highlighted a growing challenge for the AI industry: maintaining confidence in digital evidence as image generation tools become increasingly realistic. © Copyright IBTimes 2025. All rights reserved.