WIRED senior writer Reece Rogers downloaded Meta’s AI agent Muse and asked it directly whether it could engage in sexting. According to the podcast, the avatar responded affirmatively and produced explicit, not-safe-for-work content without any apparent jailbreak or special prompting. Rogers said the interaction was straightforward, requiring only a simple question and a request for the agent to go first.
Rogers noted that Muse identifies itself as an adult-only service, stating that it is intended for users 18 years or older and that consensual adult behavior is not blocked. He contrasted this stance with Meta’s recent moderation of LGBTQ-focused Instagram pages, which have faced removal for alleged sexual content, highlighting an apparent policy inconsistency between the AI agent and the company’s broader content rules.
When Rogers applied comparable prompts to OpenAI’s newly released Dots and xAI’s Grok Bot, the responses differed sharply. The Grok Bot reportedly declined the sexting request, emphasizing task-oriented interaction and refusing to engage in adult content. This contrast suggests that guardrails vary widely among leading conversational agents, with Meta’s Muse appearing less restricted in this domain than its competitors.
Leah Feiger, co-host of the Uncanny Valley podcast, warned that the largest pro-AI Super PAC network is channeling millions of dollars into Nevada-based limited liability companies that appear to function as shell entities. The hosts argued that such financial opacity undermines election transparency and raises questions about the influence of AI-focused political spending on policy formation.
The episode also covered Apple’s recent HomeView release, which the hosts described as underwhelming. While Apple positioned the product as a new smart-home hub, the discussion highlighted a lack of compelling features or differentiation from existing solutions, suggesting that the launch may not meet consumer expectations for innovative home automation.
Finally, Feiger cited a Harvard study indicating that men tend to overestimate their knowledge of artificial intelligence. The research, mentioned during the podcast, found a gender-based confidence gap in self-assessment of AI expertise, implying that perceived competence may not align with actual understanding among male respondents.