Google Earth Was The Thing Both Sides Accepted
- Johan Steyn
- 1 day ago
- 3 min read
For twenty years Google Earth belonged to nobody's argument, and a single feature launch showed how quickly that could end.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Every functioning dispute needs something both parties will accept. In courtrooms it is the record, in markets it is the audited figure, and for two decades in disagreements about land, damage and what physically exists in a place, it has been the view from above. That last one nearly stopped working last recently.
CONTEXT AND BACKGROUND
On Thursday 30 July 2026, Google added an image generation button to Google Earth, letting users zoom to any location and type a prompt to produce a photorealistic image built from the platform’s own satellite, aerial and three-dimensional data. It was withdrawn the following day. Google said it had seen geospatial professionals using the feature for a range of useful purposes but had also seen people sharing screenshots of generated imagery appearing to violate its policies, adding that people uniquely trust Google Earth for a reliable view of the world.
In the interval, researchers produced fabricated images of a nuclear plant in Iran, refugees at the Mexican border and a hospital with a bomb crater in Gaza, none of which the system refused, while journalists generated fires on Iran’s Kharg Island and a flooded United States Capitol complex. The Centre for Information Resilience warned that trust in satellite imagery had taken decades to build and could be irrevocably damaged overnight.
INSIGHT AND ANALYSIS
The strongest objection is that nothing new happened. Bellingcat’s assessment was that the change was one of speed rather than kind, since fake satellite imagery had already appeared, including images of a damaged United States base in Bahrain circulated in Iranian media, and that the integration merely streamlines a process which will make fakes proliferate. The same researcher noted a second consequence, being that governments will now more readily claim real satellite images are fake, which erodes trust from the opposite direction. That objection is correct, and it strengthens the concern rather than answering it. What made Google Earth useful was never its resolution. It was that the picture was not anybody’s submission. Once a reference layer contains a generation function, the reference stops being neutral in principle even when it remains accurate in practice, and the party who dislikes what an image shows acquires an argument they did not previously have.
IMPLICATIONS
South Africa is unusually exposed, because so many of our arguments are arguments about physical fact. Land restitution turns on what stood where and when. Municipal valuation disputes, mining rehabilitation obligations, agricultural insurance claims, construction compliance and environmental enforcement all rest on imagery serving as the neutral witness. In the United States, where the question has been litigated for longer, legal analysis of property claims notes that courts have begun applying evidentiary rules more stringently to satellite imagery because digital modification has become easier, with close attention to date markers. Organisations that rely on such imagery should expect authentication challenges they have not previously faced, and should be establishing provenance at the point of capture rather than assembling it under pressure at the point of dispute. Watermarking will not carry this weight on its own. Google’s images carried SynthID marks, which are embedded in the pixels and survive the screenshotting and cropping that destroy ordinary metadata, and the harm still spread, because the mark is invisible and confirms itself only to someone who thinks to run the check. A control that requires the audience to suspect first is not protecting the audience that does not suspect.
CLOSING TAKEAWAY
Google withdrew the feature within roughly a day, which was the right decision and a fast one. What cannot be withdrawn is the demonstration. Everybody now knows the reference layer can be forged from inside the reference layer, and that knowledge does its work whether or not the button returns. The question for South African institutions is not how to detect fabricated imagery, which is a technical problem with technical answers. It is what we will check against when the shared picture becomes something each party can produce for itself. Twenty years of accumulated neutrality is not easily rebuilt, and a country that agrees on very little can less afford than most to lose the few things it still agrees on.
Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments