Google has suspended a satellite image-generation tool integrated into Google Earth just days after its launch, bowing to pressure from disinformation researchers and verification experts who warned the feature could be weaponised to spread falsehoods at scale. The technology giant announced the rollback on August 1, citing the need to implement stronger safeguards against misuse.
The "create image" feature, released on July 30, allowed users to generate realistic satellite and aerial photographs by typing descriptions into Google Earth. Powered by Google's Nano Banana 2 image-generation system, the tool could synthesise images in seconds based on the platform's existing satellite, aerial and three-dimensional mapping databases. Google had promoted it as a means for professionals to visualise historical events, draft property development plans, and assist with urban planning. For geospatial specialists and legitimate users, the company suggested it offered genuine productivity value.
Yet within hours of deployment, digital investigations experts and fact-checking organisations sounded alarms about the implications. The concern centred on a paradox: Google Earth has spent two decades establishing itself as a trusted, reliable reference point for the global community—a resource journalists, researchers, and intelligence analysts depend upon to verify claims about conflict zones, natural disasters, and other critical events. By embedding an AI image generator directly into this platform, critics argued, Google had essentially handed bad actors the tools to erode that very trust.
During testing on July 31, international news organisations successfully used the feature to fabricate deeply concerning scenarios: a terrorist explosion in Paris, an explosion at a nuclear facility in Iran, and bomb craters in Russia. The tool also generated images of purported Islamic State training camps in Syria. Beyond geopolitical flashpoints, the system manufactured plausible false imagery of floods across Bangladesh, migrant encampments at the US-Mexico border, and a suspicious vehicle warehouse in Georgia—the latter resonating with false claims promoted by US President Donald Trump regarding alleged election fraud in that state. Each of these fabrications, had they circulated unchecked, could have sparked panic, influenced public opinion, or triggered diplomatic incidents.
Brady Africk, a satellite imagery research analyst at the American Enterprise Institute, warned that such a tool would make it dramatically easier to produce "convincing fake satellite imagery that can spread quickly online and mislead the public." He emphasised that artificial satellite images erode institutional faith in geospatial data generally, substantially complicating the work of journalists and research organisations who depend on authentic satellite photography to document human rights abuses, environmental destruction, and conflict-zone realities. The damage to verification ecosystems extends far beyond any single false image.
Google had incorporated SynthID digital watermarks—invisible identifiers embedded by the company's AI systems—into generated images as a mitigation measure. The company also referenced its AI policy, which explicitly prohibits "misinformation, misrepresentation or misleading activities." However, disinformation researchers noted that invisible watermarks offer minimal protection against deliberate falsification, particularly when bad actors can strip or obscure them before spreading images across social media platforms where context is easily lost.
The rollback decision reflects a broader anxiety within the geospatial verification community. The Centre for Information Resilience, a London-based organisation focused on digital integrity, described the feature as "irresponsible," with executive director Ross Burley warning that satellite imagery credibility has taken decades to establish and "could be irrevocably damaged overnight." GeoConfirmed, a volunteer-led open-source intelligence project tracking visual evidence from global conflicts, reported that users were already exploiting the tool to falsify military positions, mislabel buildings as schools, and conduct experiments in information manipulation—all within the narrow window between launch and suspension.
Historical precedent underscores the stakes. In 2023, an artificial image falsely depicting an explosion at the Pentagon briefly destabilised financial markets as investors reacted to the apparent attack. Earlier, during the US-Iran conflict, an AI-manipulated satellite photograph purporting to show a destroyed American military base accumulated millions of views across social platforms before being debunked. These incidents demonstrate that fabricated geospatial imagery, once released, can inflict real economic and geopolitical damage regardless of eventual fact-checking.
For Southeast Asian readers and policy makers, the incident carries particular relevance. The region has experienced sustained disinformation campaigns related to territorial disputes in the South China Sea, border tensions, and political transitions. Satellite imagery has proven invaluable for independent verification of military buildups, environmental degradation, and other geopolitical developments. A tool capable of generating convincing fake satellite photographs could undermine regional stability by enabling actors to manufacture evidence of provocative actions that did not occur, potentially triggering miscalculation or conflict escalation.
Google's decision to pause the feature suggests that the company recognises the tension between innovation and societal risk. The platform's statement indicated that development will continue, with the company working to strengthen guardrails before any relaunch. This implies that Google views the problem as one of implementation detail rather than fundamental design flaw—a position some researchers dispute. They contend that the core risk of weaponised geospatial fakes is inherent to the technology itself and cannot be fully mitigated through watermarks or policy language alone.
The episode also highlights the growing challenge facing technology companies as generative AI capabilities become more powerful and accessible. Google Earth's integration of image generation represents a logical extension of AI capabilities, yet the episode reveals how rapidly such innovations can create systemic vulnerabilities. The platform's unique position as a trusted reference source made it a particularly consequential venue for such a tool, but the underlying technology could migrate to other mapping platforms or be developed independently by other companies less responsive to criticism.
Looking ahead, the rollback may establish a cautionary precedent, encouraging technology developers to consult with disinformation researchers and verification experts before deploying similar capabilities in high-stakes information environments. Alternatively, it may be perceived as temporary restraint that precedes a more sophisticated relaunch. Either way, the incident underscores that as artificial intelligence becomes embedded in critical information infrastructure—including the mapping and satellite imagery systems that support journalism, research, and international relations—governance frameworks must evolve in tandem with the technology itself.
