The following is an edited transcript of a panel held at the Ludlow Festival of Ideas. The chair, the broadcaster Ruth Okonjo, was joined by Dr Felix Marchetti, a historian of photography; Priya Lindqvist, who analyses images for a digital forensics laboratory; and Tomasz Hale, a philosopher who writes about knowledge and trust.
Ruth: Let me begin with the obvious worry. Software can now produce videos of people saying things they never said, and the results are often indistinguishable from the real thing. Felix, is this genuinely new? Felix: The tools are new; the temptation is not. Photographs were being doctored almost as soon as photography was invented. In the Soviet Union under Stalin, officials who had fallen from favour were quietly removed from official pictures, as if they had never stood beside the leader at all. What has changed is cost and scale. Faking an image once required a skilled technician and a darkroom. Today it requires a laptop and an afternoon.
Ruth: Priya, you spend your working life trying to catch these fakes. Can it still be done? Priya: Often, yes, but I'd be wary of anyone who promises that it always will be. Early fakes had telltale flaws – people who never blinked, earrings that didn't match, light falling from two directions at once. Each time we publish a method of detection, the people making the software learn from it. It's an arms race, and in arms races the defenders rarely get to rest. That's why many of us are moving from asking 'Is this fake?' to asking 'Where did this come from?' Some cameras and editing programs can now attach a kind of digital record to an image, showing when and how it was made and changed.
Ruth: Tomasz, as a philosopher, what do you think is really at stake? Tomasz: Most of what any of us believes, we believe on testimony. I have never been to Antarctica, but I'm confident it exists, because I trust maps, teachers and the people who have been there. Video seemed to offer an exception – a kind of evidence that didn't depend on trusting anyone's word. Deepfakes puncture that illusion. But I'd add that it was always partly an illusion. Footage could be edited, cropped or shown without context long before any of this.
Ruth: So is the danger that we'll believe things that are false? Tomasz: That's the danger everyone talks about. I'm more worried about the opposite. If any recording might be fake, then anyone caught on camera doing something embarrassing can simply claim it was fabricated. Some legal scholars call this the 'liar's dividend'. The real casualty may not be credulity but trust itself. A society in which nobody believes anything is not more sceptical in any healthy sense; it is simply easier to manipulate, because people fall back on believing whatever suits them.
Felix: History suggests the same. Forgery tends to flourish when institutions are weak. When people trust a newspaper, a court or a museum, a clever fake does limited damage, because they can check it against sources they rely on. When those sources are distrusted, the fake doesn't have to be clever at all. Priya: Which is why I'm a little uneasy when the conversation becomes purely technical. Detection tools help, but they are no substitute for reputation. A video posted by an anonymous account and the same video published by an organisation with a record of correcting its mistakes are not the same object, even if every pixel is identical.
Ruth: Let me press you all for something practical. What should an ordinary viewer do? Priya: Slow down. The most effective fakes are designed to make you feel something – outrage, usually – before you think. If a clip makes you furious, that's precisely the moment to ask who posted it first. Tomasz: And accept a certain humility. 'I don't know yet' is a perfectly respectable position, and an underrated one. Felix: I'd add something more hopeful. Every new medium has produced a panic about deception, and every time people have gradually developed new habits of reading it. We learnt not to believe everything printed in a pamphlet. We'll learn this too – though the learning, I suspect, will not be painless.
Ruth: So trust is not dead, merely relocated? Tomasz: Relocated is a good word. We used to trust the image. Increasingly, we'll have to trust the chain of people and institutions that stand behind it. That's more demanding, but it may also be more honest.