
I suspect that if we could figure out a way to make open weight models without dangerous capabilities — eg tamper resistant architectures or machine unlearning — Anthropic would be in favor of doing so (so would I!).
Sara Hooker@sarahookr
Between the lines, Anthropic is saying “I’m okay with open weights as long as the model does not present high performance.” “Release what you want as long as not a direct alternative to proprietary models.”
English

