For years, Google Earth has served as a vital instrument for journalists, humanitarians, and researchers, providing high-resolution satellite imagery that has been essential in documenting war crimes, human rights abuses, and military conflicts in some of the world’s most inaccessible regions. By offering a transparent, verifiable window into global events, the platform has acted as a critical counterweight to state-sponsored misinformation, helping to authenticate reports of atrocities in places like Sudan, Gaza, and Iran. However, this established reputation for reliability was recently threatened when Google introduced an experimental feature allowing users to superimpose AI-generated imagery directly onto its mapping interface.
The feature, dubbed “Nano Banana 2,” invited users to zoom into any global location and generate custom, AI-produced visuals via text prompts. While Google intended the tool to assist with urban planning or historical visualization, the launch triggered an immediate and intense backlash. Critics and digital forensics experts warned that the ease with which users could generate hyper-realistic, fake satellite images—such as fictional nuclear plants or scenes of civilian catastrophes—created a massive risk for the spread of disinformation. By making it trivial to create deceptive screenshots that appear to be grounded in real-world geography, critics argued that Google had inadvertently handed bad actors a weapon to undermine the credibility of satellite-based evidence.
The potential for misuse was demonstrated almost instantly. Experts like Henk Van Ess utilized the tool to produce convincing, fake satellite images of fatal accidents and unauthorized military infrastructure, proving that the system could be exploited despite Google’s stated safety policies. Nathaniel Raymond, head of the Humanitarian Research Lab at Yale University, expressed shock at the rollout, noting that the feature threatened to render the work of open-source investigators “a million times harder.” For those who rely on legitimate satellite data to hold governments accountable, the existence of such a tool—even with digital watermarks—threatens to erode the public’s baseline trust in visual evidence, which is already under siege by a rising tide of AI-generated content.
In response to the mounting criticism and the rapid circulation of deceptive screenshots, Google suspended the feature on July 31, just one day after it became widely available. The company defended its initial intent, emphasizing that the AI images were watermarked to identify their synthetic nature and that they were never integrated into the primary Google Earth map view. Nevertheless, the decision to roll back the tool was an acknowledgment of the danger posed by making synthetic manipulation accessible on a platform that functions as a piece of “critical information infrastructure.” Google stated it would work to implement “stronger guardrails” before considering any future release of such capabilities.
The controversy has sparked a broader debate about the ethics of integrating generative AI into platforms that the public relies on for objective truth. Sam Gregory, executive director of the non-profit WITNESS, argued that while satellite imagery has long been susceptible to manual manipulation, Google’s integration of AI essentially condoned the act of fabricating evidence. In a digital ecosystem already drowning in hyper-realistic fakes, experts are questioning whether companies are prioritizing technological novelty over the preservation of reality. The ease of creating “ground truth” imagery through AI prompts creates a climate where every piece of satellite data could face skepticism, potentially shielding perpetrators of real-world violence from scrutiny.
Ultimately, the brief life of the “Nano Banana 2” feature has left experts calling for a significant shift in how tech giants approach product development. There is growing concern that the “move fast and break things” philosophy is incompatible with tools that serve as public utilities for human rights and historical record-keeping. Critics like Raymond and Gregory are urging Google to be more transparent about its plans and are pushing for a fundamental re-evaluation of whether AI integration is necessary for every product. For now, the suspension of the feature offers a reprieve, but it leaves open a fundamental question: at what cost are tech companies willing to sacrifice the integrity of the world’s most vital mapping resources in the pursuit of AI dominance?