Deterministic software produces the same output whenever it receives the same input under the same conditions. Unlike a generative model, it follows defined rules rather than sampling variable responses. Examples include input validation, calculations, and fixed routing logic; results can still differ if external data or system state changes.
In cybersecurity, predictable behavior makes software easier to test, audit, and enforce consistently. Defenders can use deterministic checks for access controls, policy enforcement, and known patterns, while reserving AI for tasks that require judgment. In attacks, adversaries may exploit flaws in rule-based code or supply inputs that trigger an unsafe but repeatable outcome. In AI workflows, moving predictable steps to deterministic code can also reduce unnecessary model calls, cost, and variability.