OpenAI details defenses against adversarial distillation
OpenAI says adversarial distillation lets bad actors copy a model's capabilities without its safeguards, and it describes how it detects and disrupts these attempts. In an internal test, the average attack success rate for recovering protected reasoning fell from 54.0% against GPT-6 Astra to 0.2% against GPT-6.2 Astra with updated safeguards. OpenAI says its weekly enforcement actions against these campaigns have ranged from thousands to tens of thousands of accounts. The company says it briefed Congress in February on DeepSeek's attempts to distill U.S. frontier models, and in September it attributed a core part of an extraction campaign to individuals associated with Moonshot AI.