Rate limit
An enforced budget on the most actions of a given kind an agent may take per unit of time (messages per minute, orders per hour, records modified per run). It converts a wrong action from an unbounded flood into a single, catchable event, and is the control the Knight Capital unbounded-order loss lacked (see Topic 3.5).
Defined in 2 GAGE programs, which carry 2 distinct definitions of it. The wording above is taught in Certified AI Governance Professional (CAIGP).
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
An enforced budget on the most actions of a given kind an agent may take per unit of time (messages per minute, orders per hour, records modified per run). It converts a wrong action from an unbounded flood into a single, catchable event, and is the control the Knight Capital unbounded-order loss lacked (see Topic 3.5).
A runtime control that caps how many times a given action (typically a tool call) can occur within a defined window, catching abnormal frequency regardless of whether any single instance of the action looked dangerous on its own. One of the MGF's own two named examples of a runtime control (MGF p.34).
Where it is taught
The exact lessons this term appears in. The first module of every program is free with a free account.
- The Control Typology · Technical Controls and Threat Modeling, Certified Agentic AI Governance Professional (CAAGP)
- Permissioning an agent: tools, budgets, and the actions it may never take alone · Agents Under Command, Certified AI Governance Professional (CAIGP)
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.