The story’s score and claims

Comments on No. 40 of 118 · OpenAI News · edition of Saturday 21 November 2026

Combatting adversarial distillation

Summary

OpenAI describes how it detects and disrupts adversarial distillation, where bad actors train a model on another model's outputs to copy its capabilities without its safeguards.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules