Algorithmic debt is a nonstandard term for the technical, operational, and oversight costs that accumulate when AI systems grow faster than an organization can control them. It is similar to technical debt, but it focuses on AI-specific problems such as duplicated models, unmanaged prompts, excessive model calls, weak review processes, and unclear ownership of outputs and data.
In cyber security, algorithmic debt matters because it can hide risk. Shadow AI may bypass logging, procurement, and security review, making it harder to know what data is being sent to models or which workflows are consuming resources. Over time, poorly governed AI can increase attack surface, complicate incident response, and create audit and compliance gaps. Defenses include maintaining an inventory of approved tools, tracking usage by user and project, limiting high-risk automation, and requiring human review for sensitive workflows.