Login Get a Demo
01 Glossary · Governance and risk

What Is an AI Risk Assessment?

What an AI risk assessment produces, the six steps to run one, and how it differs from an impact assessment and a fundamental rights impact assessment.

The short answer

What is an AI risk assessment?

Last updated
Read time
4 min read
From
Cranium AI
Also called
AIRA
Share on LinkedIn

An AI risk assessment is a structured evaluation of what could go wrong with a specific AI system or use case, how likely each harm is and how severe it would be, so the organization can decide whether and how to proceed.

It is one step inside AI risk management, repeated whenever the system changes.

Key takeaways
  • The EU AI Act defines risk as the probability of harm combined with the severity of that harm (Article 3(2)).
  • ISO/IEC 42001 gives it a clause of its own: 6.1.2, AI risk assessment, followed by risk treatment in 6.1.3.
  • The output is a decision with an owner: proceed, proceed with conditions, or stop.
  • A risk assessment asks what could go wrong. An impact assessment asks who is affected, and how.
02 How to run

How do you run an AI risk assessment?

The sequence below follows ISO/IEC 23894 and Article 9(2) of the EU AI Act, with NIST AI RMF references for each step.

  1. Set scope and criteria

    Name the system, its intended purpose and the scale you will score on. ISO/IEC 23894 calls this defining risk criteria (clause 6.3.4).

  2. Identify the risks

    List known and reasonably foreseeable risks to health, safety and fundamental rights, including from foreseeable misuse (Article 9(2)). Add security, legal and third-party risks.

  3. Analyze likelihood and severity

    Score each risk on one scale. NIST’s Playbook describes the common method as impact multiplied by likelihood, on a scale as simple as red, amber, green.

  4. Evaluate against tolerance

    Compare each score with your documented risk tolerance (MAP 1.5) and rank treatment by impact, likelihood and available resources (MANAGE 1.2).

  5. Decide and treat

    Mitigate, transfer, avoid or accept (MANAGE 1.3). Record the residual risk and the person who accepted it.

  6. Reassess

    Repeat on a schedule and when the model, data, vendor or use changes. Article 9 describes a continuous, iterative process, not a one-time review.

03 Compare

Risk assessment vs. impact assessment vs. FRIA

Three assessments that are often confused. They ask different questions and have different legal weight.

AI risk assessmentAI impact assessmentFundamental rights impact assessment
Core questionWhat could go wrong, how likely and how severe?Who is affected, and how, for better or worse?How could this deployment harm the rights of the people it affects?
ReferenceISO/IEC 42001 clause 6.1.2; ISO/IEC 23894ISO/IEC 42001 clause 6.1.4; ISO/IEC 42005:2025EU AI Act, Article 27
Legal weightVoluntary in the standards; part of the Article 9 duty for high-risk providersVoluntary guidanceMandatory for certain deployers of high-risk AI
OutputScored risk register and treatment planDocumented impacts on individuals, groups and societyResults notified to the market surveillance authority
04 Keep going

Related reading.

05 Questions

Frequently asked questions.

Straight answers, kept current. Last updated .

Ask us anything →
What risks should an AI risk assessment cover?

Use NIST’s seven characteristics of trustworthy AI as the prompt list: valid and reliable, safe, secure and resilient, accountable and transparent, explainable and interpretable, privacy-enhanced, and fair with harmful bias managed. Then add the risks that arrive with third parties, which GOVERN 6.1 and MAP 4.1 address: models, data and software you did not build.

Who must carry out an AI risk assessment under the EU AI Act?

Providers of high-risk AI systems, as part of the risk management system in Article 9, which must identify, estimate and evaluate risks and adopt targeted measures. Testing against predefined metrics has to happen before the system is placed on the market (Article 9(8)). Certain deployers have a separate duty, the fundamental rights impact assessment in Article 27.

Is a vendor questionnaire an AI risk assessment?

No. A questionnaire collects a vendor’s statements. An assessment scores the risk of your use of that vendor’s AI, using those statements as one input alongside documentation, test results and contract terms. NIST AI RMF GOVERN 6.1 asks for policies and procedures that address AI risks associated with third-party entities.

06 See it for yourself

From the definition
to the evidence.

See how Cranium AI discovers, observes, governs, secures and proves every AI model, agent and vendor in your enterprise.