Token-based usage is a billing model where an AI service charges for the amount of text processed, usually measured in tokens or similar units. A token may be a word, part of a word, or a punctuation mark, so a single prompt and response can consume many tokens. This model is common in large language models and API-driven AI tools.
In security operations, token-based billing matters because usage can scale quickly and unpredictably. A buggy agent, repeated retries, or malicious prompt abuse can generate large volumes of requests and create unexpected cost spikes. Defenders reduce that risk with quotas, rate limits, budget alerts, logging, and approval workflows tied to owners and use cases. Monitoring token spend also helps detect shadow AI and overactive automation before it becomes a governance or financial problem.