Until the AI can master the art of creating "footage of someone's phone if they were in the crowd of the speech in this other video", then we can't even trust that.