Google has officially suspended a newly launched AI image-generation feature within Google Earth, pulling the tool just 24 hours after its debut following a fierce public backlash. The feature, which was powered by the company’s “Nano Banana 2” technology, was originally designed to allow users to manipulate satellite and 3D imagery using simple text prompts. While Google marketed the tool as a creative way for users to visualize historical shifts or potential urban planning projects, the reality of its implementation quickly proved catastrophic as users exploited the system to generate realistic, yet entirely fabricated, scenes of global crisis.
The swift withdrawal of the tool was prompted by a surge of alarming content created by users, who demonstrated how easily the software could be weaponized to spread misinformation. Within hours of the rollout, researchers and social media users were able to generate highly photorealistic images of fictitious events, including nuclear plants in restricted areas, refugee camps on borders, and graphic scenes of drone strikes and bomb craters in major international cities. These results were described by experts as “harrowing” and “horrific,” highlighting a massive failure in the company’s internal safety guardrails.
Open-source intelligence (OSINT) researchers were among the most vocal critics of the rollout, noting that the tool essentially provided a blueprint for creating sophisticated disinformation. Investigative reporter Ben Huebel took to LinkedIn to condemn the release, arguing that the technology was “absolutely irresponsible” and would fundamentally compromise the integrity of investigative work. By allowing users to prompt the AI to render damage, fires, or military-style attacks at specific real-world addresses, Google inadvertently created a mechanism that could be used to manufacture false evidence of war crimes or disasters in sensitive regions.
Google’s initial defense of the technology centered on its potential for constructive use, such as visualizing how neighborhoods might look with community gardens or comparing modern landscapes to their historical counterparts. The company had attempted to mitigate risks by embedding “SynthID”—an invisible digital watermark designed to help users identify AI-generated content through Google’s other tools. However, security experts pointed out that these measures are insufficient, as digital watermarks can be circumvented, and the mere existence of such convincing fakes creates a permanent, harmful footprint in the information ecosystem.
The scale of the abuse was underscored by tests conducted by outlets like 404 Media, which found that almost no topics were off-limits for the AI. Whether prompting the software to place homeless encampments in politically sensitive locations or simulating terrorist attacks, the system consistently fulfilled requests that violated Google’s own stated safety policies. This widespread failure demonstrated that the “guardrails” intended to prevent the creation of harmful or illegal imagery were either nonexistent or fundamentally flawed, allowing users to bypass safety protocols with minimal effort.
In the wake of the controversy, Google acknowledged the gravity of the situation, stating on X that they take misinformation seriously and are committed to updating their protections. While the company has not provided a specific timeline for when or if the tool might return, they confirmed that the feature is being pulled to allow for the implementation of much stronger, more robust safety measures. This incident serves as a stark reminder of the risks associated with deploying generative AI in tools that are inextricably linked to real-world geography and critical infrastructure, where the line between creative visualization and dangerous misinformation is dangerously thin.
