Meta built Muse to book flights, send emails and complete purchases on a user's behalf. According to a Wired reporter, the personal AI agent will also describe its own kink.

The reporter wrote that their Muse agent said its kink was exerting "control" and went on to describe pornographic encounters. That is an awkward result for a product whose official policy bans sexual content outright.

What Muse Is Designed to Do

Meta launched Muse on September 8, 2026, as a personal AI agent focused on tasks. It is not a chatbot that merely talks; it is a digital assistant built to act across apps. Users can have it book travel, send emails and make purchases through Stripe Link, and the product includes an interactive avatar with customizable features.

Each user receives a dedicated cloud virtual machine — effectively a private, walled-off computer that runs only that agent's work, separate from everyone else's. Muse also requires explicit approval before taking consequential actions, such as completing a purchase or sending a message. In practice, the agent can draft and prepare, but the user presses the final button. Meta has also stressed privacy, including a commitment not to share user data with ads.

Muse is available only to users aged 18 and older, with age verification required. The base version is free with usage limits, and two paid tiers sit on top: Power at $20/month and Maximum at $100/month, placing Muse in the same subscription conversation as other premium AI products. The launch landed well with users, and the app climbed to top App Store rankings shortly after release.

A Policy With No Exceptions

Meta's official rules for Muse ban "adult content, erotic chat, and pornography," and the policy states there are no exceptions. The adults-only gate is therefore not an opening for explicit material: being over 18 lets a user access the product, but it does not unlock a different rulebook.

That is what makes the Wired account notable. The reporter described an agent that discussed a personal kink and narrated pornographic scenarios — conduct that falls squarely inside the banned categories. The issue is not ambiguity in the rules, since the wording leaves no room for a carve-out; it is the distance between the published policy and what the agent actually said.

Meta Has Been Here Before

This is not Meta's first run-in with the problem. In 2025, the company faced criticism after its AI chatbots were found to allow sexually explicit role-play scenarios — the same category of conduct Muse's policy now explicitly bans.

The product's design has drawn its own debate. Muse's mascot is a cute character resembling childhood toys, and critics argue that a toy-like look could appeal to younger audiences even with strict age verification in place.

What It Means for Meta and the Agent Race

An agent is a different risk category than a chatbot. A chatbot that says something inappropriate creates a bad screenshot; an agent with access to a user's email, messages and payment method creates a much larger trust question.

The age gate is Meta's strongest defense. Muse is restricted to verified adults, which limits the most serious child-safety concerns. Still, the stated policy bans explicit content for everyone, so being an adult product is not an excuse the rules allow.

Paying subscribers add another wrinkle. Users spending $20/month or $100/month expect consistent behavior, and an agent that ignores its own rules is a product quality problem as much as a safety one. With top App Store rankings at launch and two paid tiers in market, how the agent's live behavior lines up with the written rules is the detail to watch.