Skip to main content

Navigation

Human Risk

Simulations

Phishing Simulations Smishing Simulations
Vishing Simulations Soon

Catalogue

Security Awareness Privacy & Compliance AI & LLM Security Real-World Incidents Application Security API Security Git & Repository Security Cloud Security
Microlearning drills

Features

Interactive 3D Training Cloud LMS SCORM & LTI Browser Extension Integrations All features

Company

Partners About Us Blog Data Sheet PDF

Legal

Security & Compliance Privacy Policy Terms of Service
English Українська
Book a Demo

Every exercise is indexed by name, CWE, OWASP, MITRE ATT&CK, CIS, NIST CSF, GDPR and EU AI Act reference.

Try

ISO/IEC 42001 security awareness training

ISO/IEC 42001 is the certifiable standard for an AI management system. Clauses 7.2 and 7.3 ask you to prove that the people who build and run your AI know what they are doing.

Five-to-ten-minute 3D exercises mapped to each ISO/IEC 42001 duty, with completion evidence your auditor can export.

Book a Demo See the exercises

At a glance

Standard
ISO/IEC 42001:2023, first edition, December 2023
Training clauses
7.2 Competence and 7.3 Awareness
Assurance
Third-party certification audits, not fines
EU AI Act
Not a harmonised standard: no presumption of conformity

One path for the people behind your AI

AI Systems Governance & Risk Training: eight courses and two hands-on exercises, assigned again every 12 months.

Three OWASP courses
LLM applications, agentic applications and MCP, for the teams that build and wire up AI.
Two practice exercises
A deepfake voice call drill and a generative AI data-handling task.
Records per person
PDF certificates, a tenant CSV, the compliance report page, and Vanta or Drata sync.
The AI Systems Governance & Risk Training learning path in RansomLeak: AI Basics for Everyone, three OWASP courses, AI at Work, three optional policy courses, a deepfake audio drill, a safe generative AI exercise and EU AI Act Compliance

The competence loop an auditor walks through

Clause 7.2 follows the same steps in every ISO management system. The last two are where AI changes the content.

  1. 1 7.2 a Define the competence Decide what each role that affects AI performance needs to know.
  2. 2 7.2 b-c Close the gaps Train, mentor or reassign, then check that it worked.
  3. 3 7.2 d Keep the evidence Documented information an auditor can open and date.
  4. 4 7.3 Know the AI policy Everyone in scope knows the policy and their part in it.
  5. 5 Annex A Roles, people, impact Named AI roles, the human resources behind them, and impact assessment.

The path recurs every 12 months, so each surveillance audit finds a fresh record.

Where people appear in ISO/IEC 42001

Where people show up in the standard, clause by clause.

01 Clause 5: leadership and the AI policy

Top management sets the AI policy and assigns responsibilities for the AIMS.

That policy is what Clause 7.3 later expects every person in scope to know about. Training is how you show the policy reached people, not just the intranet.

02 Clause 6: AI risk and AI system impact assessment

Clause 6 asks for an AI risk assessment and an AI system impact assessment covering effects on individuals, groups and society.

The people running those assessments need to understand how AI systems fail. The OWASP courses give them a shared vocabulary for prompt injection, excessive agency and tool poisoning.

03 Clause 7.2: competence

You decide what competence each role needs, make sure people have it through education, training or experience, and check that the action worked.

You keep documented information as evidence. In practice the auditor samples a few people in AI roles and asks to see their records.

04 Clause 7.3: awareness

Everyone doing work under your control should know:

  • that the AI policy exists
  • how their work contributes to the AIMS
  • what happens when they do not follow it

This reaches beyond engineers to anyone who uses AI tools at work.

05 Annex A: roles, resources and impact

Annex A is a normative list of reference controls, with implementation guidance in Annex B. Several of them depend on people.

  • Roles and responsibilities for AI are defined and allocated.
  • Human resources, including their competences, are documented as a resource for AI systems.
  • AI system impact assessment is a defined process with named owners.
06 Clauses 9 and 10: audit and improvement

Internal audit and management review check whether the AIMS works. Nonconformities need corrective action.

A missing training record for someone in an AI role is a typical minor nonconformity. Recurring assignment stops the gap from opening again next year.

07 How the path fits an ISO 42001 programme

The AI Systems Governance & Risk Training path is aimed at teams that build or run AI systems. It has eight required items and three optional courses:

  • AI Basics for Everyone and AI at Work for shared grounding.
  • OWASP Top 10 for LLM Applications, OWASP Top 10 for Agentic Applications and OWASP MCP Top 10 for the build side.
  • The deepfake audio detection drill and the safe generative AI exercise for practice.
  • EU AI Act Compliance for the legal context.
  • Optional: Security Policies & Your Role, Protecting Sensitive Information, Incident Reporting.

Clause 7.2 is role-based, so assign the path to the people your AIMS scope names. Staff who only use AI tools can take AI Basics for Everyone on its own for Clause 7.3 awareness.

The path is assigned again every 12 months. Each completion produces a per-user PDF certificate and a row in the tenant CSV export.

The compliance report page shows completion across the tenant, and Vanta or Drata sync sends the same records to your compliance platform.

Controls that rest on trained people

The AIMS controls that only work if the people behind them are trained.

AI roles and responsibilities

Annex A expects roles for AI to be defined and allocated. A role on an org chart is not competence.

The learning path gives each person in an AI role a dated record of the courses behind it, which is what links Annex A to Clause 7.2.

AI risk: LLM, agent and MCP threats

Risk owners cannot assess threats they have never seen. The three OWASP courses cover:

  • the OWASP Top 10 for LLM Applications
  • the OWASP Top 10 for Agentic Applications
  • the OWASP MCP Top 10

Each module is built around a scenario, so engineers practise spotting the flaw instead of reading a list.

Impact assessment literacy

An AI system impact assessment asks how a system could affect people. AI Basics for Everyone and AI at Work give non-engineers enough grounding to take part in that discussion.

The EU AI Act Compliance course adds the legal side for teams whose systems reach the EU.

Data handling around AI tools

An AI policy usually says which tools are approved and what may go into them. In the Safe GenAI Usage exercise a deadline pushes the learner to paste a client email, account number and budget into a free consumer chatbot, and they practise choosing the approved tool and cleaning the prompt first.

Protecting Sensitive Information is available as an optional course for roles that handle personal or client data.

Synthetic media and social engineering

AI also changes the attacks your staff face. In the deepfake audio drill a caller who sounds like the CEO, with matching caller ID, wants a wire sent before the end of the day.

The learner applies a challenge phrase, a callback on a separate channel and the two-person rule before any money moves.

Incident reporting and the AI policy

Two optional courses support Clause 7.3 directly. Security Policies & Your Role covers what a policy asks of each person, and Incident Reporting covers what to report and how.

Make them required for roles where your AI policy says so.

Recommended exercises

Attack-side practice for engineers, everyday judgement for everyone who prompts.

What certification looks like in practice

No regulator and no fines. The pressure comes from auditors and customers.

AWS: accredited certification for four AI services

In November 2024 AWS announced ISO/IEC 42001 accredited certification for Amazon Bedrock, Amazon Q Business, Amazon Textract and Amazon Transcribe. The certifier was Schellman, accredited by ANAB.

AWS reported its first surveillance audit, completed in November 2025, closed with no findings. Certification is not a one-off: the auditor comes back.

Microsoft: Copilot products in scope

Microsoft lists Microsoft Copilot, GitHub Copilot, Copilot Studio, Security Copilot, Microsoft Foundry and others as in scope for its ISO/IEC 42001 certification.

Its compliance page makes a point worth repeating to your own team: customers can use the certificate in their assessment, but they still need an assessor to evaluate their own controls.

ISO/IEC 42006: rules for the auditors

ISO/IEC 42006, published in July 2025, sets extra requirements for bodies that audit and certify an AIMS against ISO/IEC 42001. It adds to ISO/IEC 17021-1.

For you this means auditors are themselves held to a competence standard, and they will expect the same discipline in your Clause 7 records.

Further reading

Related glossary terms

References

Give your AIMS auditor a training record that holds up

Role-specific AI courses, practice drills and dated evidence on a yearly cycle.

Book a Demo Play a free exercise

Frequently asked questions

What is ISO/IEC 42001 security awareness training?

Competence and awareness evidence for an AI management system, from prompt hygiene to OWASP risks in agents and MCP servers.

Does ISO/IEC 42001 require training?

Not by that word alone. Clause 7.2 requires you to determine the competence of people whose work affects AI performance and ensure they have it through education, training or experience. Clause 7.3 requires awareness of the AI policy.

For most organisations, training is the practical way to meet both, and the records are what an auditor checks.

Who needs ISO 42001 training?

Anyone in the scope of your AIMS whose work affects the AI systems: developers, data scientists, product owners, people who operate AI tools and the risk owners who assess them.

Everyone else in scope needs the lighter Clause 7.3 awareness: knowing the AI policy and their part in it.

Is ISO 42001 certification mandatory?

No. It is a voluntary standard. Organisations certify because customers ask for it in procurement or because it gives structure to AI governance they need anyway.

There are no fines. The consequence of weak training is an audit nonconformity that you must correct.

Does ISO 42001 certification mean we comply with the EU AI Act?

No. Under the AI Act, only harmonised standards referenced in the Official Journal give a presumption of conformity. ISO/IEC 42001 is not one of them.

The first harmonised standard for the Act, a quality management standard for high-risk AI providers, was referenced in October 2026. An AIMS still helps, because it gives you the risk, documentation and competence processes the Act also expects. See our EU AI Act page for the Article 4 literacy duty.

How does ISO 42001 relate to ISO 27001?

Both use the same harmonized structure, so Clauses 4 to 10 line up and you can run them as one integrated system. ISO 27001 covers information security. ISO 42001 covers how AI systems are developed, provided and used responsibly.

Training records can serve both if each course is mapped to the clauses it supports.

How often should ISO 42001 training happen?

The standard sets no frequency. You have to show that competence is current and that you evaluated whether training worked.

A yearly cycle lines up with surveillance audits, which is why the learning path recurs every 12 months.