Token pricing is the amount charged for each unit of model input or output, usually billed separately for prompts and generated text. In practice, this means every request has a variable cost based on how much text is sent to the model and how much it returns. Longer prompts, large document uploads, and verbose answers all increase spend.
In cyber security, token pricing matters because cost shapes behavior. Cheap tokens can encourage broader AI use, including sensitive workflows such as code review, document analysis, and internal search. That raises the stakes for access control, data classification, and approval rules. Attackers can also exploit pricing indirectly by sending oversized prompts, forcing repeated calls, or provoking long responses to create denial-of-wallet conditions. Defenders reduce this risk with quotas, rate limits, usage alerts, output caps, and careful restrictions on what data an AI system may process.