Joachim Schaeffer: OpenAI’s distillation disclosure cites our reasoning-extraction work
Joachim Schaeffer @JSchaeff3rX
Why it matters
First-hand: the researchers say OpenAI confirmed the attack paths they reported.
Summary
Schaeffer says OpenAI’s disclosure of the reasoning-extraction campaign cites his group’s work, that OpenAI confirmed the attack paths they reported and that their findings helped accelerate mitigations. Full update PDF: stolen-thoughts.com/stolen_thoughts_update.pdf.
Archived text
OpenAI just disclosed a reasoning-extraction campaign and cited our work. It says it confirmed the attack paths we reported and that our findings helped accelerate mitigations. https://openai.com/index/disrupting-a-coordinated-model-distillation-campaign/
Our full update here: https://stolen-thoughts.com/stolen_thoughts_update.pdf