OpenAI has described a layered approach to adversarial distillation that combines model training, safeguards, detection and enforcement, the company said.[1] OpenAI said the approach is meant to protect against capability extraction, meaning attempts by others to pull out what its models can do.[1] OpenAI says its layered approach to adversarial distillation combines model training, safeguards, detection, and enforcement to protect against capability extraction.[1]