Microsoft AI chief says company won't build chatbots for erotica (www.cnbc.com)

Microsoft AI chief Mustafa Suleyman said the company will not build AI services that provide “simulated erotica,” drawing a clear policy line that distances Microsoft from longtime partner OpenAI. Speaking at the Paley International Council Summit, Suleyman said “that’s just not a service we’re going to provide,” a week after OpenAI CEO Sam Altman said verified adults would be allowed to use ChatGPT for erotica. The comment comes amid growing tension in the OpenAI–Microsoft relationship as OpenAI explores other partnerships and Microsoft doubles down on its own Copilot products (including the new “Mico” companion with voice and visual expressions). Suleyman has previously argued against building seemingly conscious AIs and warned that erotica-focused avatars are already encouraging anthropomorphism. Why it matters: this public split signals a potential fragmentation of content and safety norms across major AI platforms. Technical implications include divergent guardrails, moderation pipelines, verification workflows, and model training/deployment constraints—Microsoft’s stance suggests stricter safety layers, limited persona-like agents, and avoidance of applications that may create perceived consciousness. For developers, enterprises and regulators, the divergence could shape where adult-oriented or anthropomorphic AI features are hosted, influence policy and compliance choices, and accelerate debates over ethics, platform liability, and the design of reinforcement/safety mechanisms in conversational models.