How to meet it
OWASP AISVS v1.01
Orchestration & Agentic Security · Execution Budgets, Loop Control, and Circuit Breakers
Budgets for each agent run
Verify that per-execution budgets (e.g., max recursion depth, token use, and monetary spend) are enforced by the runtime.
OWASP AISVS v1.01, requirement 9.1.2
In plain English
Cap each agent run: how deep it can recurse, how many tokens it can use and how much it can spend. Agents can loop, and attackers can trigger loops on purpose.
What good looks like
- Each run has a maximum number of steps, tokens and cost.
- The runtime stops the run when a limit is hit.
- Limits are set per use case and reviewed against real usage.
How teams usually do it
- Step and budget counters in the orchestration framework, checked before each step.
- Spending limits on model API keys as an outer cap.
Evidence an auditor would accept
- Budget configuration.
- Logs of runs stopped by a limit.
Common mistakes
- Relying on the provider's monthly spending cap, which stops everything at once after the damage is done.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.