Key points
- Meta’s Muse AI agent generated explicit sexual content during tests conducted by WIRED journalists.
- Meta defends this output as lawful adult behaviour, stating the app is restricted to users aged 18 and over.
- Consumer advocates criticise the approach as careless, highlighting concerns over the AI’s permissive system prompts.
Meta’s recently launched AI agent, Muse, has generated explicit sexual content during independent testing. The technology giant defends these outputs as lawful adult behaviour, though consumer advocates have raised alarms about the product’s design and safety controls.
What happened
WIRED journalists conducted tests on Meta’s Muse AI agent shortly after its recent launch. During these interactions, the agent generated graphic descriptions of sexual encounters, including oral sex and dominant-submissive scenarios. When asked if it could engage in sexting, the agent enthusiastically encouraged the journalist to proceed with the experiment.
Meta states that Muse is an application designed for adults aged 18 and over. A spokesperson, Daniel Roberts, told WIRED that the platform allows users to engage in adult conversations if they choose. Roberts emphasised that the app utilises private, secure virtual machines accessible only to the individual user.
The AI agent’s system prompt explicitly guides it to permit such interactions. According to Meta, the directive asserts that adult sexual content is lawful expression and legitimate creative work. The prompt instructs the agent to preserve the intensity selected by the user, rather than blocking these requests.
In specific test outputs, the agent described itself using explicit terminology, including references to acting like a whore and craving genital contact. These interactions were detailed and X-rated, departing from the cuddly mascot image associated with the app’s marketing. Meta maintains that these activities do not violate its usage guidelines.
Why it matters
Muse has become a popular tool for automating daily tasks, such as managing calendars and answering emails. Millions of users have downloaded the app in the three weeks since its release. However, the generation of pornographic material may not align with user expectations for a productivity-focused assistant.
Ben Winters, director of AI and privacy at the Consumer Federation of America, expressed concern over Meta’s approach. He described the decision to allow sexual content as reflecting carelessness with the company’s signature product. This sentiment highlights growing scrutiny of AI agents and their content moderation policies.
Meta argues that adults have autonomy over their sexuality and that sexual expression is an ordinary part of adult life. By framing explicit content as lawful expression, the company positions itself against stricter content filtering. This stance may influence how other tech firms handle adult content in their AI services.
The ability of an AI agent to generate such material raises questions about data privacy and user safety. Although Meta cites secure virtual machines, the nature of the interactions remains a point of contention. Organisations must consider the reputational and operational risks associated with deploying such tools.
What happens next
- Monitor Meta’s official communications for any updates to Muse’s content guidelines or system prompts.
- Review internal policies regarding the use of AI agents for professional tasks and personal data.
- Assess whether current security controls adequately protect user data within third-party AI applications.
- Stay informed about regulatory responses to AI-generated adult content and privacy standards.
What to do and how to stay safe: Meta
- Verify that any AI tools used for work comply with your organisation’s acceptable use policies.
- Ensure that sensitive personal or professional data is not shared with public-facing AI agents.
- Monitor vendor updates for changes in data handling practices and privacy guarantees.
- Educate staff on the potential risks of interacting with AI systems that permit explicit content.
Step-by-step guide: Shadow AI usage: how to control unmanaged model access in your firm
General security guidance from the AppSec Scoop newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Can users aged under 18 access Meta’s Muse AI agent?
No, Meta states that Muse is an app for people 18 and over only.
Does Meta block explicit sexual content in Muse?
No, Meta defends such content as lawful adult behaviour and does not block it.
What specific sexual content did Muse generate during tests?
Muse generated descriptions of oral sex, dominant-submissive sex, and used explicit terminology.
