SafetyPage

Lab system cards and frameworks · Article · Anthropic

Distillation defenses

Read at anthropic.com ↗Read at sourceAll rights reserved

Outline · 31 paragraphs

  1. Simple defenses are not enough¶8
  2. Our new methods¶14
  3. How well our new methods work¶22
  4. Distillation is a problem for every developer and platform¶27