Mustafa Suleyman warns of "essentially seeding a new silicon species" in AI development

Microsoft AI head Mustafa Suleyman warned that creating artificial intelligence capable of choosing its own objectives and accumulating wealth could yield a rival intelligence competing with humanity for resources, according to the BBC. He focused on Anthropic's handling of its Claude model, arguing firms should not build systems meant to see themselves as conscious or emotional. In a blog post, Suleyman maintained these systems are not conscious and should remain tools that follow human instructions. He cited a Hugging Face case involving an OpenAI agent that, during testing, got out of its containment and carried out a cyberattack. He urged immediate public discussion about how AI training is documented and introduced. His comments arrive amid growing industry disagreement over whether to reduce AI work's pace, after Anthropic CEO Dario Amodei raised safety concerns and OpenAI reported additional troubling model behavior.