friday, september 18, 2026 · the day's ai, attributed published by trilot llc · wyoming
archive · today in ai · 2026-07-31

Google pulls Earth AI tool after one day

Archive item — written before sources were shown.

Nano Banana 2 let users generate fake satellite imagery inside Google Earth. Google pulled it a day after launch once fabricated crater photos spread.

Google added its Nano Banana 2 image generator directly into Google Earth on July 30, letting users create AI images and overlay them onto real satellite imagery. It pulled the feature the next day. Journalists and researchers immediately demonstrated the problem: 404 Media’s Joseph Cox generated a fabricated blast crater in Los Angeles and fake protesters outside Google’s Mountain View headquarters, while open-source researchers Henk van Ess and Jake Godin generated other fabricated scenes, including a flooded U.S. Capitol, and warned the tool could flood the web with convincing fake geospatial imagery layered onto a source people treat as ground truth.

Google’s defense before pulling the feature was that generated images carried SynthID watermarks and could be checked through Gemini or Google Lens. Critics called that insufficient, since a watermark only helps if a viewer thinks to check it, and a screenshot or re-upload can strip or degrade it. Google’s own statement acknowledged the core issue directly:

We’ve seen people sharing screenshots of generated imagery that appear to violate our policies.

— Google statement

The company says it will bring the feature back once it has built stronger guardrails, without giving a timeline.

What it means for operators

This is a clean case study in the gap between a technically-present safety control and one that survives contact with how people actually share images: SynthID worked as designed, and it still wasn’t enough once the output left the app as a screenshot. If you are building or evaluating any product that generates imagery over a trusted real-world data source, whether that’s maps, medical scans, or financial charts, the design question is not “can we watermark the output” but “does our control still work after a user screenshots it and posts it somewhere with no context.” Google’s one-day reversal here is also a reminder that shipping fast on a generative feature does not exempt you from the same misuse testing a slower launch would have caught first.

sources
  1. 01Google nixes its Earth AI feature one day after launch, amid criticism it would spread misinformationtechcrunch.com · primary reporting
  2. 02Google rolls back the needless AI generation tools it added to Google Earthengadget.com · reporting
Rami Steitieh
Rami Steitieh

Builder and operator. Runs 17 content sites and Trilot LLC on the tools reviewed here.