The models will eventually defeat themselves. As AI begins training on its own output there will be a feedback loop and eventually it will devolve in to nonsense.