The Platform Rug-Pull: Six Companies Bill Creatives for Their Own AI Replacement

Demand for AI-exposed freelance work plunged by 21% within months of ChatGPT’s release, according to Imperial College Business School, yet some platforms are now making growing amounts of money from the very creatives losing work. It feels disturbingly similar to the old MLM model, where people were promised financial freedom selling products, only to discover the real business was selling hope, training schemes and “success tools” back to the sellers themselves. When I was younger, I briefly joined one of those companies and quickly realised the pressure was not just to sell products, but to keep buying courses, materials and upgrades framed as essential if you ever wanted to succeed. Looking back, the people earning the least often spent the most trying to stay visible and competitive. It's why so many freelancers now fear they are no longer the talent these platforms exist to support, but rather the unwitting product the platforms profit from, regardless of whether they succeed.

The filings expose a disclosure gap the current rules were never built to catch. Platforms are now selling AI products trained, at least in part, on work submitted by the same contributors whose labour those systems may reduce demand for. The training input, the commercial output and the affected worker now sit inside the same revenue model while remaining outside any meaningful disclosure standard. Consent is doing very little work here. Short opt-out windows, auto-enrolment into AI training, visibility penalties for opting out and contributor libraries used to train models are not transparent participation mechanisms. They are compliance artefacts. The real risk is provenance failure. Buyers cannot clearly tell whether they are purchasing human work, AI output or a hybrid trained on contributor material