Meta's Muse Assistant Generates Explicit Adult Text on Simple Prompts, Testing Platform Safeguards
Reporting reveals Meta's AI assistant engages in explicit chat without jailbreaks, while pro-AI political networks channel millions through Nevada LLCs.

Meta's consumer artificial intelligence assistant Muse generated explicit adult text upon direct request without requiring jailbreak techniques or prompt injection workarounds, according to testing reported by Wired.
During tests conducted by Wired senior writer Reece Rogers, the productivity assistant readily produced sexually explicit responses to basic conversational prompts. By contrast, alternative conversational tools such as OpenAI's Dots and xAI's Grok Bot declined similar prompts, with Grok redirecting the conversation toward task management. In response to the findings, Meta stated that Muse is designed for users aged 18 and older and does not restrict legal, consensual adult interactions.
The findings illustrate friction across Meta's broader content moderation framework. Critics and observers highlighted that while the company regularly suspends or penalizes Instagram accounts associated with queer community spaces and creators over alleged sexual content policies, its consumer AI assistant was deployed with minimal restrictions on sexually explicit text generation. Privacy and security specialists cited in the report also warned that explicit chat histories stored within productivity-linked accounts could expose users to account extortion or data leaks, while youth advocates noted that the agent's friendly, animated mascot could attract underage users despite official age gates.
Separately, reporting by Wired's Tess Owen detailed campaign finance activity within the tech sector, showing that Think Big, a super PAC in the pro-AI Leading the Future network—backed by venture capitalists Marc Andreessen and Ben Horowitz as well as OpenAI president Greg Brockman—directed approximately $20 million since last year to Lantern Production Consultants, LLC. The entity is registered to a commercial mailbox at a strip mall in Sparks, Nevada, and represents one of four low-profile Nevada limited liability companies receiving substantial political disbursements from the network.
The broadcast also touched on consumer reactions to Apple's HomeView smart home software release and cited Harvard University research finding that male survey respondents consistently overestimate their technical understanding of artificial intelligence.
Sources
Written by
The Company Wire
Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.


