A fabricated Tom Holland and Zendaya street clip shows the deepfake's new target: the boring moment

The facts we have are thin, so we will be honest about the limits: a clip circulating online shows Tom Holland and a pregnant Zendaya getting out of an SUV on a New York street, and the detection service AI or Not identifies it as AI-generated. Nothing in the material tells us who made it, how far it spread, or what it was for.

Even so, the case is worth flagging, because it marks a shift in what synthetic video is being used for. The deepfakes that shaped public debate were built around outrage: a politician saying something incendiary, a celebrity in a compromising scene. Those fakes carry their own alarm bell — the content is implausible enough that someone checks. A couple walking to a car carries no alarm bell at all. There is nothing to fact-check because nothing is being claimed.

That is the harder problem. Mundane synthetic footage does not spread as a lie; it spreads as filler — pulled into gossip aggregators, tabloid embeds and recommendation feeds where nobody was ever going to ask for a source. It quietly manufactures a record of events that did not happen, and around real people whose private circumstances are being narrated for them.

Our reading: detection tools like the one that flagged this clip are useful, but they are a race we will keep losing at the margin, because they judge the pixels after the fact. The durable answer is provenance — signed capture, verifiable chain of custody, platforms that surface where a file came from rather than whether a classifier dislikes it. As generation gets cheaper and better, the scarce, valuable thing stops being the image and becomes the proof of where it originated. That infrastructure is buildable now, and it is the unglamorous work that decides whether the next decade of synthetic media is a nuisance or a corrosion.