ai_security▌

← feed

NOTABLE

OpenAI details adversarial-distillation campaigns and reports cutting reasoning-extraction attack success from 54% to 0.2%

OpenAI's first public write-up on adversarial distillation treats unauthorized extraction of protected reasoning traces as a cyber/bio capability-transfer risk and describes layered classifiers, API-level request/response protections and disruptions of coordinated extraction campaigns; in an internal adaptive evaluation with automated red-teaming, average success at recovering protected reasoning fell from 54.0% against GPT-6 Astra's original safeguards to 0.2% against GPT-6.2 Astra with the updated safeguards.

released 2026-11-20 · ingested 2026-11-21T00:55Z

Sources

Primary source

Cite this

APA

AI Security Index. (2026). OpenAI details adversarial-distillation campaigns and reports cutting reasoning-extraction attack success from 54% to 0.2%. AI Security Index. https://openai.com/index/combatting-adversarial-distillation/

BibTeX

@misc{aisec-2026-11-20-openai-adversarial-distillation-defenses,
  title  = {OpenAI details adversarial-distillation campaigns and reports cutting reasoning-extraction attack success from 54% to 0.2%},
  year   = {2026},
  howpublished = {\url{https://openai.com/index/combatting-adversarial-distillation/}},
  note   = {AI Security Index item 2026-11-20-openai-adversarial-distillation-defenses}
}