Know Their Reach · PART 1 · QUESTIONS 01-05 Score: 1 point per Yes, consistently

Yes, consistently Partly Not yet

01 Can you name the owner and business purpose of every

agent you run?

02 Can you map every data source, API, tool, and MCP server

each agent can reach?

03 Do you know which identity and permissions an agent

uses for each action it takes?

04 Can you follow one agent task step by step across every

system it touches?

05 Can you explain afterward why an agent picked a tool or

took an action?

Control Their Actions · PART 2 · QUESTIONS 06-10 Score: 1 point per Yes, consistently

Yes, consistently Partly Not yet

06 Can you let an agent read a system while denying write,

delete, and execute?

07 Can policy weigh who is asking, what the agent is for, and

how sensitive the data is before an action runs?

08 Do high-impact or irreversible actions wait for a human

decision?

09 Can you stop a manipulated agent from misusing tools it

is allowed to use?

10 Can you pause or disconnect an agent the moment it acts

outside its purpose?

Know Their Reach Yes answers / 5 Control Their Actions Yes answers / 5

Count your Yes, consistently answers per part. Partly and Not yet count as zero. A control that works sometimes is a gap.

4–5 You can answer for your agents. Book a session to benchmark against enterprises running agent control at scale.

2–3 Coverage is uneven. Agents are doing things you can only partly account for, and these gaps compound as agent count grows.

0–1 You can't currently answer for what your agents reach or do. This is where a session helps most.

A I S E C U R I T Y C H E C K L I S T S E R I E S

The Agent Control Ten Ten questions every organization should be able to answer about what AI agents can reach and what they are

allowed to do. Mark one answer per question, reflecting how consistently it works today.

Bring your two scores to a session

Reach and control gaps call for different fixes.

Talk to an AI security expert

Scan or tap to learn more

checkpoint.com/ai-security/

https://www.checkpoint.com/ai-security/ https://www.checkpoint.com/ai-security/

agent_control: question_01: Off question_02: Off question_03: Off question_04: Off question_05: Off question_06: Off question_07: Off question_08: Off question_09: Off question_10: Off score: know_their_reach: control_their_actions: