Anthropic staff say AI could kill everyone. But its safety rule keeps getting weaker

Anthropic staff say AI could kill everyone. But its safety rule keeps getting weaker