normative claim
AI model weights should be openly distributed as the default. Capability concentrated in a handful of closed labs is a greater structural harm than any hypothetical misuse enabled by open release. The burden of proof rests on restriction, not on distribution.
norm_open_weights_distribution