Red teaming is the adversarial half of responsible AI: the OWASP category checklist, a program with intake and cadence, and a defense stack measured end to end. This module sets up the operating model and the runnable harness, and hardens the reporting habit that keeps a safety number honest.