Google has abruptly halted a major artificial intelligence update to its Google Earth platform just one day after its release. The decision to roll back the tool follows a wave of criticism from researchers and investigative journalists who demonstrated how easily the technology could be exploited to create realistic, yet entirely fabricated, satellite and aerial imagery.
What happened
On Thursday, Google introduced a generative AI feature powered by its “Nano Banana 2” technology. The tool was designed to allow users to modify Google Earth’s 3D and satellite landscapes by simply typing text prompts. However, by Friday, the company announced it would be pulling the feature after users began sharing disturbing and misleading content generated by the system.
During the short window the tool was active, users were able to produce photorealistic images of global catastrophes that never occurred. Reports indicated that individuals used the interface to “place” nuclear power plants in sensitive geopolitical regions, generate fake refugee camps at international borders, and depict fatal traffic accidents in major European cities. Investigative outlets like 404 Media further demonstrated the tool’s risks by generating images of bomb craters in Los Angeles and fabricating protests outside of Google’s own corporate offices.
Context
Google originally marketed the generative tool as a way for users to visualize urban development and historical changes. In their initial announcement, the company suggested that the AI could help residents imagine community gardens in vacant lots or see what their neighborhoods might have looked like a century ago. The intent was to provide a creative utility for real estate planning and historical education.
Despite these benign intentions, experts in open-source intelligence (OSINT) quickly identified fatal flaws in the system’s safeguards. Researchers like Henk van Ess and investigative reporter Ben Huebel pointed out that the AI did not appear to have significant “off-limits” prompts. This lack of oversight allowed for the creation of imagery that closely mimicked the visuals coming out of actual conflict zones, such as those in Ukraine or the Middle East. While Google has previously touted its AI detection and watermarking efforts, critics argued that these indicators are often easy to bypass, making the generated images dangerous tools for disinformation.
Why it matters
The removal of the feature highlights the ongoing tension between rapid AI innovation and the responsibility to prevent digital harm. For the OSINT community, which relies on satellite imagery to verify human rights abuses and military movements, the introduction of “hallucinated” satellite photos represents a significant threat. If fake aerial footage becomes indistinguishable from reality, the process of verifying international events becomes exponentially more difficult.
Furthermore, this incident serves as a cautionary tale for tech giants attempting to integrate generative AI into platforms that are widely viewed as authoritative sources of truth. Because Google Earth is used globally for research and education, the stakes for accuracy are much higher than in standard AI art generators. The swift rollback suggests that Google recognized the potential for its platform to be weaponized in information warfare, emphasizing that even the most advanced AI requires robust guardrails before being released to the general public.