If an AI video defames someone, who should the victim actually sue, the person who prompted it, the model provider, or the platform that kept it up?

This feels like a question we're going to have to take seriously pretty soon. We've already seen cases where AI generated likenesses are used in ways people never agreed to, from unauthorized ads to digital replicas that go beyond what someone thought they had consented to. I've been paying more attention to AI video tools lately, including platforms like PixVerse. What's interesting is how accessible AI video creation is becoming. That opens up a lot of creative possibilities, but it also makes me wonder how responsibility should work when someone uses these tools to create something that causes real harm. The current legal landscape is a patchwork: right of publicity, defamation, copyright, and platform safe harbors can all point toward different parties. The person who prompted the video might not have the resources to compensate a victim. The model provider may point to its terms of service, while the platform hosting the video may have its own legal protections depending on the jurisdiction. I'm not a lawyer, but I'm curious what a workable liability framework would look like. Should responsibility depend on who created the content, who distributed it, and how each party responded after being notified? If the only party with enough resources to compensate the victim is the platform or model provider, should that affect how liability is assigned? And how do we avoid creating a chilling effect on legitimate creative tools? For those who follow AI policy or platform governance, what's the fairest way to divide responsibility when an AI-generated video causes real reputational harm?