The Cost of Governance: Measuring Overhead in Toggleable LLM Agent Safety Layers
7,504 live API calls. Three providers. Five governance configurations. The full safety stack adds 0.2 to 0.35 milliseconds. Token overhead: zero.
Finding: A full agent-safety stack — classifier, filter, router, audit logger — adds between 0.198 and 0.355 milliseconds per call. Token overhead is zero. Across 7,504 live calls to three providers, total experiment cost was $4.37.
Paper abstract and link coming soon. arXiv: preprint pending.
Last updated: March 1, 2026