Understanding safety requires separating accidents from hazardous states. An accident occurs when a system in a hazardous state meets adverse environmental conditions, turning risk into loss. Designers can only influence the system, not the environment, so the safest approach is to prevent the system from entering hazardous states altogether. Ignoring this distinction leaves safety to chance, for example, automotive brakes that can fail under heat.

A recent commercial flight touched down with under 30 minutes of fuel, triggering alarm despite the reserve being technically sufficient. In aviation, landing with less than that threshold defines a flight fuel hazardous state; only severe weather could then cause a crash. Regulators therefore mandate a minimum fuel buffer to keep aircraft out of that risky zone. Airlines track fuel continuously, letting pilots trim routes.

Maintaining safety constraints becomes a dynamic control problem involving feedback, predictive models, and corrective actions. When any of these three control structure components falters—poor sensor data, inaccurate forecasts, or insufficient actuation—the system may slip into a hazardous state. Aerospace and autonomous‑vehicle teams add redundant sensors and simulations. Recognizing and fixing such deficiencies before an accident occurs offers a more proactive path to reliability.