Mustafa Suleyman argues Anthropic treats AI as if it were human. He warns that such an approach could produce technology humanity cannot control. He spoke with the BBC about the risks of human-like AI.