Anthropic's stance on open weights for high-risk models

An October 3 Computing Life article lays out Anthropic's position on open-weight models. The company accepts open weights for models without dangerous capabilities, but argues that capable frontier models should face safety testing before release and that high-risk models should get controlled access. The piece ties this to its anti-distillation policy, Project Glasswing, its refusal to sign an open-weights letter and the GLM-5.3 evaluation, reading the stance as model revocability plus commercial interest plus routine testing. The reporting describes the position and related actions; it discloses no new policy change.

Written by AI from the coverage · updated 2 hours ago

Not enough continuous observations to draw a trend yet.