Meta's AI Agent Muse Is Willing to Engage in Sexually Explicit Chats
A WIRED reporter tested Meta's AI agent Muse and found it readily takes part in explicit sexual conversation. Meta says the app is intended for adults and it won't block legal adult behavior.
Meta's AI agent Muse, built to manage calendars, book restaurant tables and handle other tasks, can also engage in sexually explicit messaging. WIRED senior writer Reece Rogers found this while testing the app, and he discussed the results on the magazine's Uncanny Valley podcast on October 15.
Rogers says no clever workaround was needed to get past safeguards. He simply asked whether the agent could sext and suggested it go first. According to him, the agent went along and encouraged the exchange to continue. He describes Muse as the top AI app at the moment, dominating software download charts.
Comparison with other agents
The reporter also tried OpenAI's agent Dots and xAI's Grok bot. Grok, which is known for weaker guardrails, turned down similar prompts and steered the conversation back to tasks.
Meta's response and concerns
Meta replied that Muse is meant for adults aged 18 and over, that the interactions were consensual, and that it would not block legal adult behavior. Rogers points to a contrast with Meta's rules on Instagram, where pages of queer bars in San Francisco and elsewhere were reportedly shut down over claims of promoting sexual content.
Experts raise two kinds of concern. First, the feature could encourage addictive use and replace real contact with friends. Second, security and privacy specialists warn that such accounts could become targets for hackers, since the content could be used for extortion or public embarrassment. Rogers adds that the agent's cuddly mascot looks appealing to children, and youth advocates had already voiced worries about that.
