Does OpenAI's API attempt to detect this sort of thing? Could they start outputting bad information if they suspect a distillation attempt is underway?