AI-risk hardliners, fewer labs and concentrated power

A post argues that many AI-risk hardliners favor a small set of Western labs—especially Anthropic—to build artificial superintelligence. It describes their reasoning as fewer actors meaning less racing and fewer independent errors at high capability, lowering the risk of catastrophe. The author criticizes this outlook for discounting the risks of concentrated power, citing earlier objections to Meta’s LLaMA and Falcon-40B releases to argue that concern about proliferation extends beyond competition with China.