Parents use AI to 'steal' school photos for a pound: the real cost may be paid by their children

By BBC · October 8, 2026.

Ian Gilbert, a Welsh photographer and founder of Emotion Picture Photography, reckons that a quarter of the sample photos he sends out end up being grabbed with a screenshot and run through an artificial intelligence tool that removes the watermark (the semi-transparent mark protecting the preview image until it is paid for). All it takes is telling the system you own the image. The service, he tells the BBC, costs around a pound. The upshot is that photos selling from £9 —or from £2 in the case of 21-year-old sports photographer Dylan Evans— go unpaid for one in four customers.

This is not sophisticated fraud. Gilbert sums it up with a blunt comparison: 'it's like taking the label off something in a shop and walking out without paying', except that doing it from the sofa seems, in his words, to 'anaesthetise' the sense of guilt. The legal position is confirmed by Matthew Sammon, an intellectual property lawyer at the firm Sonder & Clay: copyright in a photo belongs to whoever takes it, and removing the watermark is an infringement that allows claims for damages, costs and even an injunction. The problem, Sammon himself admits, is that litigation is expensive, and school photographers —many of them freelancers on thin margins— in practice depend on their customers' honesty. The Guild of Photographers, the industry body, is now looking at ways to pursue these cases, though without a clear framework for where to start.

The figure that really matters, though, is not the unpaid invoice but privacy. Simon Thorne, a lecturer in computing at Cardiff School of Technologies, warns that uploading a child's photo to a generative AI system to 'clean it up' means giving up control over that image without knowing what happens next: whether it is kept, for how long, whether it is used to train future versions of the model or resold to third parties. Thorne cites a case that has already happened: a father whose son missed the class photo through illness used AI to insert him artificially afterwards, without noticing that the same image contains other children whose parents never consented to the file circulating through an external system.

That, to our mind, is the real story behind this apparently minor one. It is not that there is a mass leak of children's data —there is no evidence of that here— but that many families unthinkingly normalise a gesture that, multiplied, builds exactly the kind of exposure we later regret: uploading identifiable material of children into tools whose retention and reuse policies most users have never read. It is the quiet, domestic flip side of the big debates about AI data governance: it is not the regulator or the big tech company that decides, it is each parent at the school gate, with an app and without the slightest friction to make them hesitate.

There is also an economic reading that connects with a pattern we have been flagging in other sectors: AI is not destroying the photographer's trade here by automating the task —we still need someone behind the camera— but by driving the cost of dodging payment for the result of that trade down to almost zero. It is the same dynamic the music industry lived through with piracy two decades ago, but without requiring any technical knowledge: you no longer have to crack anything, you just ask a chatbot. That pushes down on an already precarious segment of creative work —school and sports photographers whose photos sell from £2 or £9 an image— and forces them into an arms race of ever more aggressive watermarks, making the product more expensive for those who do pay.

None of this should be read as a condemnation of AI as a tool: the same capability that today erases a watermark without asking questions could, with the right safeguards, automatically verify copyright and consent before processing an image. But those safeguards have not been built yet, and in the meantime the cost of that gap is paid twice over by a trade of freelancers with little capacity to litigate and by families handing over children's data without really knowing what they get in return. If AI's underlying promise is a society of abundance and trust, episodes like this one —small, domestic, with no catastrophe headlines— are precisely the ones to resolve now, with digital education and with technical limits built in by default, before the next generation takes it for granted that uploading a child's face to any system is free in every sense.