Google has quickly withdrawn a new artificial intelligence feature for Google Earth that allowed users to generate AI visualisations on top of satellite imagery. The company cited policy violations and the potential for misinformation as reasons for the retraction.
The "create image" tool, launched on Thursday, enabled users to generate photorealistic scenes using text prompts layered onto real satellite, aerial, and 3D mapping data. Critics, including researchers and open-source intelligence experts, voiced immediate concerns that the feature could be used to fabricate convincing fake satellite images, thereby spreading disinformation and eroding trust in geospatial data. Examples shared by these experts included fabricated images of explosions in Paris, nuclear sites in Iran, and training grounds for militant groups. Google's statement on X, the platform formerly known as Twitter, acknowledged that while geospatial professionals found uses for the feature, others were sharing screenshots of generated imagery that appeared to violate their policies. The company stated it was rolling back the feature to implement "stronger guardrails" to prevent misuse.
The tool, powered by Google's Nano Banana 2 image-generation technology, allowed users to zoom into any location on Google Earth, tap a button, and input text prompts to create new visualisations. While Google stated that these generated images did not appear in the main Google Earth experience for other users and were watermarked with its SynthID digital tracking technology, concerns remained that screenshots of these fabricated images could be easily shared online. This potential for misuse was highlighted by incidents such as an AI-generated image of an explosion at the Pentagon that briefly impacted markets in 2023, and altered satellite images depicting destroyed military bases during the early stages of the US war with Iran.
The Centre for Information Resilience, a London-based organisation, described the feature as "irresponsible." Digital investigations expert Henk Van Ess noted that Google Earth, built over two decades as a trusted source of global information, had introduced a feature that "makes things up." Brady Africk, a research analyst at the American Enterprise Institute, told AFP that the update would have made it "easier to generate convincing fake satellite imagery that can spread quickly online and mislead the public," further complicating the work of journalists and researchers.
Google initially defended the feature, asserting that users could verify image authenticity through the Gemini app or Lens in Search, due to the SynthID watermark. However, the rapid backlash and demonstration of potential misuse, including generating fake nuclear sites and other falsehoods superimposed on real coordinates, led to the swift retraction. This incident follows a previous instance in February 2024 where Google paused Gemini's image generation capabilities for people after the tool produced historically inaccurate depictions.
Google's Generative AI Prohibited Use Policy outlines restrictions against generating dangerous, illegal, hateful, or harmful content. This includes content that facilitates illegal activities, violates the rights of others, or compromises the security of Google's services. The company has stated it expects users to engage with generative AI responsibly, legally, and safely. The quick reversal of the Google Earth feature underscores the challenges technology companies face in balancing the introduction of powerful generative AI tools with the imperative to prevent their misuse for disinformation and other harmful purposes. The company plans to reintroduce the feature with enhanced safeguards.