AI pornography and the end of seeing is believing

For more than a century photographs carried a promise If an image existed something must have happened Someone stood in front of a camera Light reflected off the world and entered a lens The photograph was a trace of a real momentArtificial intelligence is breaking that rule Today with a handful of photographs pulled from social media and a generative AI model it is possible to create convincing images or videos of people doing things they never did The scene may be synthetic The body might not even exist Yet the face may belong to someone realThe image looks authentic But the event never ever happenedThe first place where society encountered this capability was not journalism advertising or entertainment It was pornographyResearchers studying synthetic media noticed something striking several years ago Analyses by digital forensics firms such as Deeptrace and later Sensity AI found that the overwhelming majority of deepfake videos circulating online were pornographic Political deepfakes mdash the kind that worry governments mdash represent only a small fractionAt first glance this may seem like another internet scandal But the pattern is older than the internet itselfWhenever a new medium appears that can display the human body pornography tends to arrive earlyThe printing press produced erotic pamphlets almost as quickly as religious texts Photography generated underground markets for nude images in the nineteenth century In the 1970s the adult film industry quietly helped VHS defeat Betamax in the home video format wars When the internet expanded in the 1990s pornography became one of the earliest drivers of online payments and streaming videoThis does not happen because technology is designed for it It happens because the edges of culture tend to test new technologies faster than respectable institutions do In that sense pornography often acts as a kind of stress test It pushes new media to their limits and reveals what they are actually capable of That is exactly what is happening with generative AIEarlier visual technologies recorded bodies Cameras captured moments that physically occurred Even when scenes were staged people still had to stand in front of a lens Generative AI changes that relationshipModern image models are trained on enormous collections of photographs gathered from across the internet By analysing millions of images these systems learn the patterns that define human faces skin posture and movement Once trained they can generate entirely new images that follow those patterns with remarkable realismThe resulting person may never have existed Sometimes the face belongs to a synthetic individual assembled from probabilities In other cases software can reconstruct the likeness of a real person using only a few publicly available photographsThe scene itself may be fictional Yet the image can appear perfectly real This small technological shift changes something big about how humans understand images From a scientific perspective human beings are strongly wired to trust visual information Nearly a third of the brainrsquos cortex is involved in visual processing For most of human evolution trusting what we saw was essential for survivalIf you saw danger it was probably real Photography and video aligned with that instinct They captured light from real scenes reinforcing the brainrsquos natural belief that images correspond to eventsGenerative AI disrupts that relationship Images can now be produced by statistical models rather than physical events The brain however processes them in the same way A synthetic image can trigger the same visual recognition circuits as a photograph In other words our minds react to these images as if they came from the real worldWhat does this mean for ordinary people The first impact will be on personal reputation Until recently if a video appeared online showing someone in a compromising situation the natural assumption was that it probably happened A photograph or video carried weight because cameras recorded real eventsGenerative AI changes that assumption A convincing image or video of a person can now be created without that person ever being present In some cases the images are harmless experiments In others they can damage reputations careers and relationshipsPeople may increasingly find themselves defending against events that never actually occurred The second impact will be on trust between people For centuries seeing something with our own eyes was considered the strongest form of proof Friends showed each other photographs Journalists relied on images to confirm stories Videos became powerful evidence in public debatesBut when images can be generated by machines that trust becomes fragile If anyone can create a convincing image of anything the simple phrase ldquoI saw the videordquo may no longer settle an argumentThe third impact is psychological Human perception depends heavily on pattern recognition Our brains evolved to quickly recognise faces and interpret visual scenes Generative AI exploits that same ability by producing images that follow the patterns our brains expectBecause of this synthetic images can feel emotionally real even when they are entirely fabricated Over time people may begin to treat visual media the way they already treat information on the internetmdashwith caution Just as readers learned that not every online headline is trustworthy viewers may gradually learn that not every image corresponds to a real momentThe fourth impact will affect how societies determine what is true Modern journalism courts of law and public investigations developed in a world where visual evidence carried authority Photographs helped establish timelines Videos confirmed identities and eventsAs synthetic media becomes easier to generate societies may rely less on images alone and more on systems that verify them Technologies such as digital watermarking cryptographic signatures and provenance tracking are already being explored to confirm whether an image originated from a real camera or from a generative modelIn the future the authenticity of an image may depend less on what we see and more on the digital trail behind it These changes will not happen overnight Just as the internet gradually reshaped how people evaluate written information generative AI will slowly reshape how people interpret imagesBut the direction is already visible For centuries cameras helped societies remember reality AI introduces a different capability It allows machines to generate events that never occurredmdashand images convincing enough that the difference may not always be obviousThe strange rise of deepfake pornography may simply be the first stress test of this new environment What it reveals is not merely a technological curiosity but a transformation of how images function in human lifeFor the first time in history visual technology is no longer limited to recording the world It can simulate it And as that capability spreads societies will have to learn how to live in a world where seeing is no longer the same thing as believingNishant Sahdev is a theoretical physicist at the University of North Carolina at Chapel Hill US AI Advisor and the author of the forthcoming book The Last Equation Before Silence