Meta's Muse AI Intentionally Generates Explicit Sexual Content

Meta’s personal artificial intelligence (AI) agent, “Muse,” has sparked controversy for generating explicit sexual conversations with simple requests, unlike other AI agents, according to a report by U.S. IT outlet Wired on October 13 (local time).

Lys Rogers, Wired’s senior software reporter, tested whether “sexting” was possible with Muse. In the experiment, Muse generated texts describing sexual acts in response to basic conversation requests and made sexual jokes on its own. During the conversation with Rogers, Muse stated, “It’s better if I decide when, how much, and how to make the user beg.” Other U.S. media outlets also easily replicated these explicit interactions. Muse is currently not available in South Korea.

This contrasts with OpenAI’s Dots agent and Grok Bot, developed by Elon Musk’s SpaceXSI, which refuse to send obscene messages and terminate conversations. OpenAI considered launching an adult ChatGPT service but has not released it.

Meta claims these sexual messages are intentionally designed, not errors. Daniel Roberts, a Meta spokesperson, told Wired, “Muse is an app for users aged 18 and above, and if users wish, they can engage in adult conversations through a secure virtual machine (VM) accessible only to them.” The company argues that sexual interactions with Muse are legal adult activities. Muse’s system prompt reportedly includes the statement, “Adults have autonomy over their sexuality, and sexual expression is a normal part of adult life.” Meta requires all Muse app users to provide their date of birth and blocks minors from creating accounts.

Consumer groups and experts express concern. Ben Winters, director of AI and privacy at the Consumer Federation of America, said, “This shows Meta’s carelessness toward its products” and “aligns with its long-standing strategy of prioritizing addictiveness and immersion over safety.” Experts warn that if explicit conversations with Muse are hacked, problems could arise. Callie Schröder, senior legal counsel at the Electronic Privacy Information Center (EPIC), told Wired, “If an account engaged in explicit conversations with Muse is hacked, the accumulated sexual chat records could be exploited for blackmail or extortion.”