Google removed a new AI image-generation feature from Google Earth less than 24 hours after launch, after users produced realistic fake scenes tied to real locations and raised immediate concerns about misinformation, public safety and the reliability of satellite imagery.

The tool allowed users to type a prompt and generate photorealistic images grounded in Google Earth’s satellite, aerial and three-dimensional data. It was powered by Google’s Nano Banana 2 model and could place fictional events, buildings or damage onto recognizable locations.

Google said it paused the capability after seeing users share generated images that appeared to violate company policies. The company did not specify which images triggered the decision but said it was adding stronger safeguards before considering a relaunch.

The reversal exposes a problem that ordinary AI image generators do not create at the same scale. Google Earth is widely treated as a factual mapping and imagery service, so fabricated scenes produced inside the platform can appear more credible than images generated in a separate creative application.

Users were able to create artificial scenes involving disasters, armed conflict and other sensitive events at real-world locations. Even when the images were obviously fictional to the person generating them, screenshots could be detached from their original context and circulated as evidence of an actual event.

Google said the generated images were watermarked and did not appear in the main Google Earth experience. Those protections reduced the risk of the platform itself confusing generated content with authentic imagery, but they did not prevent users from sharing screenshots elsewhere.

Watermarks also depend on people knowing where to look and trusting the detection system. Once an image is cropped, compressed or reposted, ordinary viewers may not recognize that it was generated by AI.

The business implications extend beyond Google. Mapping platforms are used by news organizations, insurers, real-estate professionals, logistics companies, governments and emergency-response teams. Their value depends on users believing the underlying geographic information reflects reality.

A tool that can quickly generate convincing false imagery threatens that trust. Insurers could face fabricated property damage claims, investors could react to fake scenes involving factories or ports, and emergency officials could be forced to verify images before responding.

Real-estate professionals were among the intended users. Google promoted the feature as a way to visualize redevelopment plans, historical scenes and possible uses for empty land. Those applications remain commercially useful, but they require a clear separation between planning concepts and current conditions.

The same technology could help architects, municipalities and developers show how a neighborhood might look after construction. Yet a realistic visualization can become misleading when it is presented without the prompt, timestamp or AI label that explains how it was created.

Google’s decision illustrates the difficulty of adding generative AI to products built around factual information. The more closely an AI output resembles a trusted record, the greater the harm when safeguards fail.

Search engines, maps and satellite platforms carry a different responsibility than entertainment tools because users often rely on them to make decisions. Speeding an image feature to market without sufficient controls can therefore create legal, reputational and operational risks far larger than the feature’s immediate revenue potential.

The pause also shows how quickly public testing can uncover weaknesses that internal evaluations miss. Google launched the tool on Thursday and withdrew it Friday after researchers and users demonstrated how easily it could be used to create deceptive scenes.

Competitors will face the same challenge as geospatial AI expands. Satellite and mapping data can support urban planning, disaster forecasting, agriculture and infrastructure analysis, but combining those systems with unconstrained image generation creates an obvious path to manipulation.

Google now must decide whether stronger guardrails can preserve the commercial value without undermining trust in Google Earth itself. That may require blocking sensitive prompts, limiting the locations that can be altered, embedding visible labels and making generated images easier to authenticate outside the platform.

Pulling the feature after one day prevented a larger rollout problem, but it also revealed how little margin for error exists when generative AI is placed inside a product people use as a record of the physical world.

JBizNews Desk | Mountain View, California

© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.