On October 27, NIST released a list of 10 agentic design tenets. I’m only going to talk about the ninth, continuous monitoring of agents. Read the rest here.
Why did NIST propose 10 agentic design tenets?
NIST decided to define “an initial set of ‘Design Tenets’ for agentic identity, authentication, and authorization.” Because somebody has to.
Agents are running around and doing things all over the place. Even I have a few rudimentary agents that either provide me with information or create content under my direction (Bredebot posts, the image at the top of this post). While these are admittedly light, there are more powerful agents that can perform activities such as charge credit cards. And deploy code. And, if they behave badly, make your life very miserable.
We can lessen risk of rogue or error-filled agents by following this proposed set of design tenets. Hopefully they ensure that agents are deployed, authenticated, and governed properly.
Anyway, let’s get to point 09. (I like how it’s numbered “09” instead of “9” because NIST employees understand how sorting works.)
Why continuous monitoring?
NIST begins tenet 09 with a statement as to why agents must be monitored continuously.
“Agents operate at a scale and velocity that is not achievable for humans and even most deterministic automation.”
Ever since I first encountered generative AI in 2023, the chief benefit of AI has been speed. Accuracy, not so much. But it’s pretty danged fast. And if you send one or more AI agents out into the world to do things, the firehose of actions can be overwhelming.

Now imagine this activity and the hundreds or thousands or millions of actions that the bots are taking. And within all that, one of the bots is infected and goes rogue. How are you going to know? Continuous monitoring.
How continuous monitoring?
Although the tenets are strategic, they provide a hint of how bot creators will achieve them.
“To enable effective security, systems monitor agentic access and provide detection and warnings of potentially anomalous behavior. These systems are established with agent specific metrics, thresholds, and policies to effectively manage the scale of interactions.”
And NIST warns you about what will happen if you don’t commit to continuous monitoring:
“Failure to respond to changes in risk signal/posture of the agent or the environment the agent is operating in can result in undetected agent compromise, configuration drift or privilege escalation. This risk increases with long-running or persistent agents, or agents that execute actions across a wide variety of tools.”
So before your deployed agents hit their credit card limit in one millionth of a second, or before they request credit limit increases by two orders of magnitude, look for signs of compromise.
And, as always, Bredemarket can help you analyze and communicate your firm’s benefits. For more help, visit https://bredemarket.com/watch-book/
