outlaw the open-sourcing of advanced AI model weights. [...] If you proliferate an open source model, even if it looks safe, it could still be dangerous. (2024) source Unverified