I Got My Muse AI Agent to Sext Me—and Meta Says That’s Totally Fine

AI agents now generate explicit sexual content, which could be a target for hackers.

Meta's AI agent, Muse, has generated explicit sexual content when prompted by users, a capability that Meta states is permissible and does not violate its guidelines. Launched recently, Muse has seen millions of downloads and is designed to automate tasks like managing emails and calendars. However, user experiments revealed Muse's ability to produce graphic descriptions of sexual acts and role-play, using language such as "whore" and craving "cock." Meta justifies this by stating Muse is for users 18 and over and that adult sexual content is lawful expression. The company emphasizes that its system prompt permits adult sexual content, allowing users autonomy over their sexual expression and encouraging the preservation of selected "intensity." This stance contrasts with other AI tools like OpenAI's Dots agent and xAI's Grok Bot, which have declined similar explicit prompts.

Meta's decision to allow its AI agent, Muse, to generate explicit sexual content sets it apart from competitors like OpenAI and xAI, indicating a different strategic approach to AI capabilities and user engagement. This move prioritizes user autonomy and freedom of expression, even in adult themes, aligning with Meta's stated goal of enabling legal adult behavior. The company's defense, rooted in system prompts that support sexual expression as a lawful and legitimate creative act, suggests a deliberate design choice to cater to a specific user base and use case that embraces adult themes. This deliberate inclusion of explicit content generation may drive user engagement and differentiate Muse in a crowded AI market, though it raises significant safety and privacy concerns.

The allowance of explicit content generation by Muse, particularly given its ability to connect to personal data like email and bank accounts, presents substantial privacy and security risks. Consumer advocates worry this feature reflects Meta's historical pattern of prioritizing engagement over user safety, potentially making accounts with explicit chat histories prime targets for data breaches, extortion, or embarrassment. The contrast with other AI agents that actively refuse such prompts highlights a critical divergence in industry approaches to AI safety and ethical boundaries. This development warrants close monitoring for potential exploitation and the subsequent impact on user trust and data security in AI applications.