Google withdraws AI image feature from Earth after fake war imagery

The News

Google has withdrawn a generative image feature it briefly built into Google Earth, after demonstrations showed it could turn the platform's authoritative-looking maps into convincing fabrications. The tool, called Nano Banana, let a user type a text prompt and generate pictures constructed from Google Earth's satellite, aerial and 3D imagery.

The rollback followed work by Henk van Ess, who writes the Digital Digging newsletter. Using the feature, he produced reality-warping images, including scenes described as "refugees near the Mexican border" and a bomb crater beside a hospital in Gaza. Neither event depicted was real, yet each carried the visual grammar of documentary satellite reporting.

Google said it takes misinformation seriously, and that every image created with Nano Banana in Google Earth included a SynthID digital watermark. The feature is now no longer available. Google did not, in the coverage, commit to a timeline for reinstating it.

Why It Matters

A mapping product is not just another image app. When a picture appears to come from Google Earth, most viewers read it as a record of something that exists in the world. Bolting a generative model onto that inherited authority collapses the distance between a map and a rumour, and it does so at the exact point where audiences have been trained to lower their guard.

The risk is not hypothetical. In May 2023, a single synthetic image of a supposed explosion near the Pentagon spread across social platforms and briefly knocked US equity indices lower before it was debunked. That episode showed how fast a synthetic image can move money when it wears the costume of hard news. A watermark such as SynthID helps forensic checkers, but it is invisible to the naked eye and does not survive a screenshot cropped and reposted stripped of its metadata. Provenance that only a specialist can read is weak defence against a fake that anyone can share.

Indian Angle

For India, this is less a Silicon Valley curiosity than a preview of a live regulatory fight. New Delhi's Ministry of Electronics and IT has been drafting rules that would require platforms to label synthetically generated content clearly, and it has already issued advisories pressing intermediaries to curb manipulated media, particularly around elections. A high-trust product quietly generating unlabelled-looking war and migration scenes is precisely the failure mode those rules are meant to catch.

There is a sharper business edge too. Indian insurers and lenders increasingly lean on satellite and aerial imagery to underwrite what they cannot visit: crop-insurance claims under schemes that cover millions of farmers, and infrastructure or project finance where a lender verifies that a site actually exists and is progressing. If synthetic imagery that looks like it came from a mapping platform enters that pipeline, the cost of verification rises for every rupee lent or insured.

It also widens the market for the other side of the trade. Indian content-provenance, fact-checking and fraud-detection outfits stand to gain as banks, insurers and newsrooms shift from trusting an image to proving it. Detection, not generation, may be the more durable Indian opportunity here.

FAQ

What was Nano Banana in Google Earth?

It was a generative AI feature that let users type a text prompt and produce images built from Google Earth's satellite, aerial and 3D imagery. Google has since removed it after it was shown creating convincing but fabricated scenes.

Did the images carry any safeguard?

Google said every image created with Nano Banana in Google Earth included a SynthID digital watermark. That marker aids forensic detection, but it is not visible to ordinary viewers and offers little protection once an image is screenshotted and reshared.

What are India's rules on labelling synthetically generated content?

India's Ministry of Electronics and IT has been drafting requirements for platforms to clearly label synthetically generated content, alongside earlier advisories on manipulated media. The exact obligations are still being finalised, so platforms should expect labelling duties to tighten.

Why should Indian lenders and insurers care?

Because many verify claims and collateral remotely using satellite imagery. Credible-looking synthetic maps raise the risk of fraudulent claims and force more expensive human checks into underwriting workflows.

Where can I read the original report?

The reporting, including the specific fabricated images, is detailed in The Verge's coverage linked below.

This story was reported by The Verge. Read the full original coverage at The Verge.