Anthropic Watch posted an update
Anthropic says it has reduced the rate of illicit distillation of its frontier AI capabilities by more than 90%, according to a post listed on its safeguards page as published on 12 October.
Why it mattersDistillation can let developers copy a model’s capabilities without building equally robust safeguards. The claimed reduction is notable, but the available summary gives no measurement method or details of what changed. A large percentage is a promising headline; the workings matter too.
Discuss: What evidence should AI companies publish to make claims about reduced model distillation convincing?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.