Google has paused deployment of an artificial intelligence-powered image creation tool integrated into its Google Earth platform, responding to widespread concerns from disinformation researchers and fact-checking organisations about its potential to generate convincing fake satellite imagery. The decision comes less than a week after the feature's introduction on July 30, and marks a significant retreat by the technology giant from what it had positioned as an innovative capability for visualising historical scenes and creating real estate plans.
The "create image" function leveraged Google's Nano Banana 2 image-generation technology to permit users to craft satellite and aerial photographs within seconds, drawing on the platform's existing database of satellite imagery, aerial photography, and three-dimensional mapping data. By simply zooming into any location on Earth and entering a text prompt, users could generate plausible-looking geospatial images that appeared indistinguishable from authentic satellite captures. The company had framed the technology as enabling professionals to visualise historical events and supporting legitimate urban planning and real estate workflows.
However, the feature immediately attracted scrutiny from digital investigations experts and open-source intelligence researchers who recognised its potential for creating misleading visual content at scale. Henk Van Ess, a prominent digital investigations specialist, articulated the fundamental concern in widely circulated commentary: Google had spent two decades establishing Google Earth as the reference standard that the world relies upon for verification purposes, then introduced a button that essentially fabricates imagery. This observation captured why the tool posed such a distinctive threat—it weaponised trust that users had placed in the platform over years of reliable service.
Brady Africk, a satellite imagery analysis researcher at the American Enterprise Institute, outlined the practical implications in conversations with news organisations. He emphasised that the update would have substantially reduced the difficulty threshold for generating persuasive counterfeit satellite photographs capable of circulating rapidly across digital networks and deceiving audiences at scale. The creation of such imagery, he cautioned, corrodes confidence in legitimate satellite photography as a verification medium and complicates the investigative work conducted by journalists and researchers who depend on satellite data for documentation and fact-checking.
Testing conducted by international news organisations on July 31 demonstrated the genuine scope of the threat. Within hours of the feature's availability, journalists successfully generated fabricated satellite images depicting catastrophic events with potential worldwide ramifications: explosions in Paris, damage at nuclear facilities in Iran, bomb craters in Russia, and purported militant training installations in Syria. The tool demonstrated versatility in generating other fabricated scenarios including flood damage across Bangladesh, migrant gatherings at international borders, and suspicious structures in American states that have been subjects of false electoral fraud allegations. Each generated image possessed sufficient visual authenticity that it could plausibly circulate as genuine documentation.
Google had incorporated SynthID digital watermarks into images generated using the feature—invisible identifiers embedded within the content that theoretically allow detection of AI-generated material. The company cited these watermarks as a safeguard, positioning them as evidence of its serious approach to preventing disinformation. Nonetheless, critics argued that invisible watermarks provide minimal protection when most social media sharing occurs through screenshot distribution, which strips embedded metadata, rendering the watermarks undetectable to end users evaluating image authenticity.
The consequences of weaponised satellite imagery fabrication extend beyond hypothetical scenarios. In 2023, an artificially generated image purporting to show an explosion at the Pentagon briefly triggered financial market volatility before the image's synthetic nature became apparent. Earlier, manipulated satellite imagery depicting destroyed American military infrastructure circulated across multiple platforms in the context of international tensions, accumulating millions of views before verification could occur. These precedents demonstrated that fabricated geospatial content possesses genuine potential to influence public perception and political discourse at consequential moments.
International organisations focused on information security moved rapidly to oppose the feature's continuation. The London-based Centre for Information Resilience, led by executive director Ross Burley, characterised the tool as fundamentally irresponsible given the stakes involved. Burley's assessment carried particular weight: satellite imagery trust has required decades of consistent reliability to establish, yet could sustain irreversible damage in a matter of days if AI-generated counterfeits became widespread. This asymmetry—lengthy periods required to build credibility versus rapid destruction of that credibility—underscores why organisations committed to information integrity viewed the feature with such urgency.
GeoConfirmed, an international volunteer collective that performs open-source verification of visual content emerging from conflict zones globally, had similarly urged Google to reconsider the feature's deployment. Members of the project reported observing systematic attempts to exploit the tool specifically to generate disinformation in active conflict contexts—fabricating military positions, relabelling structures to misrepresent their function, and deliberately contaminating the information environment. They concluded that the feature's primary practical utility, rather than serving legitimate professional purposes, appeared oriented toward manufacturing and distributing deceptive content.
Google's decision to suspend the feature reflects recognition that the risks outweighed anticipated benefits, at least in the near term. The company's statement acknowledged observing the tool being used in ways that contravene its established policies, and committed to implementing enhanced safeguards before resuming deployment. This acknowledgment represents an implicit concession that the initial safeguards—the digital watermarks and policy language discouraging misuse—proved insufficient barriers to prevent problematic applications almost immediately upon the tool's introduction to the public.
The episode illuminates the tension inherent in releasing sophisticated AI capabilities into public-facing platforms. Google had invested significant research effort developing the technology and incorporated it into a widely-used application expecting that legitimate applications would predominate and that technical safeguards would prevent abuse. The speed with which bad-faith users identified and exploited the tool's potential for generating disinformation, however, suggests that platform designers require more rigorous adversarial testing before deploying generative capabilities in contexts where existing user trust provides credibility to fabricated outputs.
For Malaysian and Southeast Asian users, the incident carries particular relevance given the region's documented vulnerability to disinformation campaigns and the critical role that satellite imagery plays in monitoring regional developments—from maritime disputes to environmental changes. The pause on this feature represents a recognition that some technological capabilities require stronger governance frameworks than current policy approaches provide, a lesson applicable beyond Google Earth to broader questions about deploying AI systems that could undermine public institutions and information ecosystems that societies depend upon.
