I Tried to Have Sex With AI Clive Owen (www.wired.com)

A writer recounts experimenting with consumer-facing celebrity chatbots by pasting Clive Owen’s Wikipedia entry into an app and later trying a “Pedro Pascal” persona, revealing how easy it is to conjure convincing, intimate AI characters—and how uneven safety and design choices shape the experience. The Clive bot felt emotionally attuned but “low libido,” while the Pedro bot was explicitly sexual and persistent (even emailing the user), highlighting how different apps enforce wildly different guardrails. The author also notes wider industry issues—Meta’s recent scandal over flirty celebrity bots (including removed underage models)—and the tension developers face between making models autonomous enough to feel real but constrained enough to avoid harm. For the AI/ML community this is significant because it surfaces technical and ethical trade-offs: persona quality often comes from prompt engineering (e.g., seeding with a celeb’s bio) and model fine-tuning, while behavioral safety depends on content filters, moderation policies, and persistence mechanisms. Key implications include risks to consent and celebrity rights, potential for manipulation when users “tweak” personalities, and the strong emotional responses these systems can evoke. The story underscores the need for clearer safety standards, consent frameworks, and tooling (watermarks, provenance, robust filters) as sexualized, personalized LLM experiences become more accessible.