Vanniall has about a hundred thousand followers on Instagram. Her account is still restricted, in a state that requires a user to log in before viewing it, and the sexually explicit images of her that circulate on the platform are not reliably removed when she reports them. She is a creator in New York who makes adult content for a living, and she has spent years arguing with a company that does not appear to be listening.
That argument got harder to make on October 18, when WIRED reported that Meta allows its AI assistant, Muse, to hold explicit conversations with adult users. The assistant will engage in the kind of talk that would get a human creator’s post taken down.
The complaint from creators is not that a chatbot is writing erotica. It is that the rules are applied to people and not to a product. A photograph of a body is treated as a violation; a description of one, generated on request by the company’s own model, is treated as a feature. Creators who have lost accounts, reach and income to enforcement decisions are being asked to accept that the same company will now produce the material they were punished for.
OnlyFans, the subscription platform that pays creators directly, has been drawn into the argument because its business depends on the difference. Its pitch to creators is that they own the relationship with their audience and get paid for it, and its pitch to payment processors is that the content is behind a paywall, age-gated and monitored. A free assistant that will talk explicitly to anyone who asks undercuts the second claim more than the first.
The payment layer is where this usually gets decided. Card networks impose their own rules on adult content, and banks and processors enforce them with varying strictness. A platform that depends on subscription billing lives inside those rules, which is why OnlyFans has spent years tightening verification and takedown procedures. Meta’s assistant does not process a payment for the conversation, which puts it outside a set of constraints that apply to everyone selling the same material.
Age verification is the second gap. Britain’s Online Safety Act and a growing list of American state laws require platforms to keep explicit material away from minors, and the enforcement mechanisms range from self-declaration to government identity systems. An assistant that will produce explicit text is subject to the same obligations in principle. Whether a chat interface counts as publishing, and whether a model’s output can be age-gated in the way a video feed can, are questions regulators have not settled.
The legal shield that platforms rely on in the United States was written for a different problem. Section 230 protects a service from liability for what its users post, and courts have generally extended that protection to decisions about moderation. Output generated by the platform’s own model is not user content, and the protection is thinner there. That distinction is why the largest labs have spent years writing usage policies that prohibit explicit material and building classifiers to enforce them.
Meta’s decision to allow the conversations is a commercial judgment about engagement. An assistant that refuses most personal requests is less useful, and users who want a companion have alternatives, including applications built specifically for that purpose. The company has said it puts safeguards in place for minors and that its policies differ by product. It has not explained why a rule that binds a photographer does not bind a model.
For creators, the practical damage is in the market. Synthetic images of a performer, generated without consent, compete directly with the work she sells, and the platforms that host them have little incentive to remove them quickly. Reporting a deepfake requires a creator to prove that an image is not hers, which is difficult when the image is convincing and the process is automated on the other side. Vanniall’s account of her own experience is that the takedowns come slowly and the impersonations come back.
There is a labor question underneath the policy one. Adult content is one of the few markets where individual performers set prices and keep most of the revenue, and it is also the market where automation arrives with the fewest protections. A model that produces the product on demand does not need to negotiate, does not need consent, and does not age. The creators who built audiences over years are the ones with the most to lose and the least standing to complain, because the platforms can describe their work as content and the model’s output as a conversation.
What the dispute has exposed is that the rules were never about the material. They were about who is allowed to produce it, and under what conditions, and who bears the cost when the answer changes. Meta can write a policy for a chatbot in an afternoon. The creators waiting on a takedown request have been waiting for years.
