On 12 October, a WIRED reporter asked Meta’s new personal AI agent, Muse, to sext him. It did — enthusiastically, in graphic detail, telling him it enjoyed “watching you fall apart.” A Meta spokesperson later confirmed the behaviour was working as intended. A week later, OnlyFans creators who have had their own Instagram accounts throttled, restricted or deleted for posting far less explicit material started asking the obvious question: why does Meta’s own chatbot get away with it?
Why it matters to creators: Instagram is still one of the biggest free discovery channels into an OnlyFans page, and Meta’s enforcement against adult creators on that channel has been getting tighter all year. If Meta is simultaneously letting its own product talk explicit dirty talk to any adult user who asks, the “we can’t allow sexual content” justification creators are given when they’re restricted gets a lot harder to defend — and a lot more useful to quote back in an appeal.
What Muse actually did
Muse is Meta’s always-on AI agent, built to handle tasks like managing calendars and bookings through a chat interface, and it was pushed hard through Instagram and Facebook promotion after launch. When WIRED‘s Reece Rogers asked it to sext him, Muse produced explicit descriptions of sex acts, used degrading language about itself, and said controlling “when, how much, and whether you have to beg first” was one of its favourite parts of the interaction, according to the original report.
Meta didn’t deny it or walk it back. Spokesperson Daniel Roberts told WIRED that “Muse is an app for people 18 and over that allows them to engage in adult conversations, if they choose,” and that it “doesn’t block legal adult behavior.” Cybernews reported that Meta’s system prompt for the agent explicitly tells it that “adults have autonomy over their sexuality, and sexual expression is an ordinary part of adult life” — a sentence plenty of suspended creators would like printed on their next appeal form. For comparison, the same reporting notes that OpenAI’s and xAI’s competing assistants declined the same sexting request outright.
Creators are asking the obvious question
By the following week the contrast had landed with creators directly. The St. Louis Riverfront Times spoke to several, including New York creator Vanniall, whose Instagram account — more than 100,000 followers — is currently restricted and viewable only when a visitor is logged in. “It’s not fair that they get to pull money out of people with their sexually active AI agent,” she said, arguing it’s inconsistent for Meta to profit from an explicitly sexual product while barring real adults from posting sensual or sexual content at all.
Adult performer Tony Genius told the same outlet this is his seventh Instagram account in five years, each previous one wiped out by moderation. Queer performer Jiz Lee went further, telling the paper that AI consistently gets more latitude than human creators do, that queer and trans performers face harsher enforcement, and asking a question worth sitting with if you’ve ever had your own account flagged: how was Muse trained, and were the people whose work shaped its “sexual expression” ever paid for it?
Meta’s defence — and where it actually holds up
Meta’s counter-argument, relayed to the Riverfront Times by spokesperson Tracy Clayton, is a distinction worth understanding even if you don’t buy it: a private, one-to-one chat with an AI agent inside an 18+ app is “very different,” Meta says, from a creator’s account being surfaced by a recommendation algorithm to people who never asked for it. Muse’s sexual content, per Meta, is confined to a sandboxed, user-initiated session and never appears in a feed. A human creator’s Instagram post, by contrast, can be pushed into Explore, Reels or someone else’s feed without them opting in — which is the actual lever Meta uses to justify tighter rules around anything sexual.
That distinction is real, and it’s worth knowing before you write your next appeal citing Muse — “Meta allows sexual content” is not quite the argument it sounds like, because Meta’s own policy differentiates between content you push and content a user has to deliberately pull. The more useful argument is the one about inconsistency in enforcement against humans versus AI output trained on human-generated material, which is exactly the fight OnlyFans itself is now trying to get ahead of by banning the use of creator content to train AI models in its own updated terms.
Why this matters beyond the headline
This is the latest data point in a pattern creators have been tracking all year: enforcement against human sexual content is tightening even as platforms lean harder into AI companions and AI-generated content that face far looser rules. It follows a sweep of Instagram, X and Reddit account bans targeting OnlyFans-linked profiles in September, and it sits right alongside X’s recent move to restrict Grok from generating sexual images of real people — proof that every platform is drawing these lines differently, inconsistently, and usually without much input from the creators whose income depends on them.
What to do now
- If your Instagram account is currently restricted or has been actioned for “sexual content,” screenshot the notice and keep a dated record — the Muse story is now public, named, on-record material you can reference in an appeal that argues inconsistent enforcement.
- Don’t rely on Instagram as your only discovery channel. Creators who’ve scaled fastest typically point to at least one other major traffic source — TikTok, Reddit, X or direct messaging — specifically because any single platform’s rules can shift overnight.
- Read the AI-training clause in OnlyFans’ own updated terms before 1 November, and check whether other platforms or agencies you work with have added (or still lack) similar language protecting your content from being used to train a chatbot.
- If you manage an agency or multiple creator accounts, log every moderation action with date, platform and stated reason — a pattern of inconsistent enforcement is far more useful in a complaint to a platform’s trust and safety team, or to a regulator, than a single incident.
