Fallback planning is the process of preparing an alternate tool, model, service, or workflow before the preferred one becomes unavailable or is removed. In AI systems, that might mean keeping a second model ready, preserving a stable API version, or defining manual review steps if automation fails. It is a basic resilience control: when a dependency changes, the organization can continue operating instead of stopping abruptly.
In cyber security, fallback planning matters because attackers and defenders both depend on reliable systems. A forced switch, deprecation, or outage can create gaps in monitoring, prompt handling, access control, or business logic. Defenders use fallback paths to reduce downtime, preserve validation, and avoid silent behavior drift when a model is retired or updated. Good fallback planning includes testing the alternate option, documenting who can activate it, and checking that it does not weaken security assumptions.