Open weights vs. closed: An AI civil war's afoot, and the stakes are existential
TL;DR
It started as China vs. the US, but it's become a face-off between two fundamentally different ways of building LLMs. And the safety of everything is on the line. It started as China vs.
Nauti's Take
This fight is productive even if the volume is high. Open weights deliver auditability, independent safety research and price pressure on closed providers, a real opportunity for teams running their own infrastructure.
The limit is irreversibility: once weights are public nobody pulls them back, and guardrails come off in minutes. Companies handling regulated data sleep better with closed models and contracts, while teams that need control and portability win with open weights.
Sources