Google pulls Google Earth AI image tool after deepfake alarm

The interesting bit in this story is not that Google shipped an AI feature and pulled it in a day. That happens. The interesting bit is which feature, because satellite imagery has been the one visual medium that reporters, human rights investigators and open source analysts still treated as roughly trustworthy by default.

According to NPR, Google unveiled a 'create image' button in Google Earth on Thursday. Bryan Horowitz, a Google product manager, described it as 'Just zoom in to a place in Google Earth on web, tap create image, and type whatever you want to see.' A day later the company said it was rolling the tool back, telling NPR that 'we've also seen people sharing screenshots of generated imagery that appear to violate our policies. So we're rolling back this feature in Google Earth while we work on implementing stronger guardrails.'

The reaction was sharp because satellite imagery does verification work that photos and videos long ago stopped doing. Henk van Ess, an open source researcher, told NPR he tried 'refugees at the Mexican border, a nuclear plant in Iran, a crash in Amsterdam, a hospital with a bomb crater in Gaza. Nothing was refused.' Others generated Iran's Kharg Island oil terminal on fire and a flooded US Capitol. Jake Godin, a senior researcher at Bellingcat, told the outlet that 'satellite imagery has been kind of a safe bet when it comes to verifying an event because it's hard to fake.' Evan Hill, a visual forensics investigator at the Washington Post, was blunter, saying he was 'very curious how or if this idea was red teamed internally because the opportunities for abuse and disinfo are literally boundless.'

Why this matters if you don't work an investigations desk: the class of evidence journalists and NGOs use to say 'yes, something happened at this coordinate on this day' just got weaker at the margins. Google points to SynthID watermarks as a safeguard, but watermarks travel with files, not with screenshots, and the examples people were already sharing were screenshots. The honest caveat is that the reporting does not spell out what the stronger guardrails will actually be, whether the tool was red-teamed before launch, or when it comes back.

The forward-looking read is more prosaic than the panic suggests. Commercial imagery providers with a clean chain of custody, and the verification community that has spent years building tradecraft around them, become more valuable, not less, when the free consumer tool stops being trustworthy. Take the specifics as reported here rather than settled, but the direction of travel is the part worth watching.