OpenAI details layered defense against adversarial distillation

OpenAI News · rss · 2026-11-20

OpenAI published an official post on combating adversarial distillation — the extraction of model capabilities via distillation. Its layered approach spans model training, safeguards, detection, and enforcement to protect against capability extraction, offering a rare look at the company's anti-distillation strategy.

Original post →

More from Models

Models channel →