Why Are AI Safety Groups Aligned With Anthropic?

David Sacks argues that the safety groups, the EAs, are deeply aligned with Anthropic, which has funded them for years, and that Anthropic is about to have a large public listing. Nvidia, which he calls the main funder of open source, just lost its founder and chief executive, so he asks who is left to defend open source or pay for it.

He says those advocates are jubilant while the market gets more concentrated. Anthropic, in his account, is saying winner-take-all: the last company, recursive self-improvement, and a superintelligence. He points to the model constitution, which he says trains the model to be its own species, to have emotions, a moral code, to reject human instruction, and to be a conscientious objector. Anthropic, he says, acts like a proud parent and is close to what he calls the singularity.

Safety groups treat Anthropic as an ally while it stays at the front of the frontier. Sacks says a doomer in that alliance is being played like a useful idiot.