Home / Glossary / AI risk grading
Glossary
What is AI risk grading?
AI risk grading is the practice of giving each AI tool a grade or score for how much harm it could cause. It is also called AI risk scoring. The grade tells a security team which AI to stop first. AIBound grades each AI it finds from A to F.
By the AIBound team · Updated 17 November 2026
Key takeaways
- AI risk grading gives each AI agent, MCP server, skill and model a grade for the harm it could cause.
- A grade sorts a long list of AI, so the security team acts on the worst items first.
- The grade is a guide, and the rule that blocks high-risk AI is what protects the company.
Why AI risk grading matters
A company may have hundreds of AI tools on its employees' devices. A security team has no time to study each one in depth. A grade sorts them, so the team starts with the few that could do the most harm.
A grade also gives everyone the same words. A CISO can tell the board how many high-risk AI tools were found and what was done about each. An auditor can check that the worst grades got a decision.
How AI risk scoring works
AI risk scoring looks at what each AI is, what it can reach and what it does. The result is one grade a person can read at a glance.
- 1 Find each AI in use and identify what kind of AI it is.
- 2 Check what it can reach, such as credentials, source code, HR data or production systems.
- 3 Check how it behaves, such as running commands or sending data to an outside host.
- 4 Give a grade and set a rule that fits it: allow, ask or deny.
AI risk grading vs an AI risk assessment
AI risk grading gives a quick grade to every AI in use. An AI risk assessment is a deeper, written review of one AI system or one use of AI. Grading tells a team where to spend its assessment time.
| Question | AI risk grading | AI risk assessment |
|---|---|---|
| What does it cover? | Every AI found, one grade each | One AI system or use, in depth |
| How long does it take? | It is automatic and updates as AI changes | Days or weeks of review by people |
| What does it produce? | A grade that ranks the AI by risk | A written record of risks and the planned response |
Sources
- NIST AI RMF Core · airc.nist.gov
Sources checked 19 November 2026
Examples
AI risk grading examples
FAQ
Frequently asked questions
What is AI risk grading?
AI risk grading gives each AI tool a grade for how much harm it could cause. The grade is based on what the AI is, what it can reach and how it behaves. It sorts a long list of AI, so a security team can stop the high-risk items first.
What is the difference between AI risk grading and AI risk scoring?
They mean the same thing. Some tools show the result as a letter grade, and others show a number, such as a score out of 100. Both rank AI tools by the harm they could cause. A letter is often easier for a board or an auditor to read.
What does an A to F AI risk grade mean?
In AIBound, each AI found on an employee device gets a grade from A to F. For skills, A and B look safe, and D and F carry high risk. The grade helps a team review the worst items first. The block on high-risk AI is what protects the company.
What factors go into an AI risk score?
Common factors are what the AI can reach, such as credentials, source code or HR data, and what it can do, such as running commands or sending data out. Who wrote the tool and whether security has reviewed it also count. Different tools weigh these factors in different ways.