Preprint
Aug 2026
Stealing Reasoning Traces from Proprietary LLM APIs
An architectural vulnerability is identified that circumvents anti-distillation mechanisms, allowing adversaries to extract a proprietary model's reasoning, as well as proposing concrete cryptographic and system-level mitigations to secure client-side reasoning.
Alexander Panfilov, David Schmotz, I. Shumailov et al.
· 9 citations
· ⚡3