ISO/IEC 27090 AI Security RACI Template
Who is Responsible, Accountable, Consulted and Informed for AI security, from scope and threat assessment to testing, suppliers, monitoring, incidents and review. Nine roles, 32 activities, sample assignments you can change, and a check that every activity has exactly one Accountable role.
Template & sample only — a structured starting point with sample content, not legal, audit, certification or investment advice.Subscribers customise it with their own details and generate the document for their own work.
RACI preview: scope and threat assessment
R = Responsible, A = Accountable, C = Consulted, I = Informed. The full template has 32 activities across eight stages, plus role holders.
| Stage | Activity | Board / exec sponsor | CISO | AI product owner | ML engineering | AI security engineering | Data / privacy | Legal / compliance | Procurement | Internal audit |
|---|---|---|---|---|---|---|---|---|---|---|
| Scope and governance | Approve the AI security scope, objectives and risk appetite | A | R | C | - | C | C | C | - | I |
| Scope and governance | Keep the AI system inventory and AI bill of materials current (built, bought and embedded AI) | - | A | R | R | C | C | - | C | I |
| Scope and governance | Name an accountable owner for each AI system and record whether you are provider, deployer or both | I | A | R | C | C | - | C | - | I |
| Threat assessment | Data and model poisoning: training, fine-tuning and retrieval data | - | A | C | R | R | C | - | - | I |
| Threat assessment | Evasion: crafted inputs that make the model misclassify or misbehave | - | A | C | R | R | - | - | - | - |
| Threat assessment | Membership inference and model inversion: privacy leakage through the model | - | A | C | C | R | R | C | - | - |
| Threat assessment | Model theft and extraction through queries or leaked weights | - | A | C | C | R | - | C | - | - |
| Threat assessment | Prompt injection and unsafe output handling, including agents and their tools | - | A | C | C | R | - | - | - | - |
| Threat assessment | AI supply chain: third-party models, datasets, weights and AI libraries | - | A | C | C | R | - | C | R | - |
| Threat assessment | Rate each AI threat and record it in the risk register with a named owner | I | A | R | C | R | C | C | - | I |
22 more rows in the full template. Customise it with your organisation’s details and download it as Word, PDF, Markdown or CSV with Pro.
A RACI matrix for AI security, organised along the life cycle that ISO/IEC 27090 guidance follows: scope and governance, threat assessment for each AI threat theme, control design, testing, supplier assurance, monitoring, incident response and review. Nine roles, 32 activities, sample assignments you can change. The summary checks that every activity has exactly one Accountable role and someone Responsible.
Your context
How to read this matrix
RACI by life-cycle stage
One row per activity. Exactly one A per row; use A/R where the accountable role also does the work. Add rows for activities specific to your organisation.
Role holders
Who holds each role today. A role can be one person, a team or a supplier.
Loading…
TL;DRWhat this template gives you
4 points
- A RACI answers one question per activity: who does it, who owns the outcome, who is asked first and who is told after.
- Nine roles: board or executive sponsor, CISO, AI product owner, ML engineering, AI security engineering, data and privacy, legal and compliance, procurement, internal audit.
- Thirty-two activities across eight life-cycle stages, including a threat assessment row for each AI threat theme (poisoning, evasion, privacy leakage, model theft, prompt injection, supply chain).
- The summary flags activities with no Accountable or more than one, activities nobody is Responsible for, and the roles carrying the most load.
Frequently asked questions
Does ISO/IEC 27090 require a RACI?
ISO/IEC 27090 is guidance, not a set of certifiable requirements, so it does not require any particular document. A RACI is a practical way to show that each AI security activity has an owner, which ISO/IEC 27001 and ISO/IEC 42001 auditors will look for when AI security sits inside those management systems.
Why exactly one Accountable per activity?
If two roles are accountable, neither is. One Accountable role owns the outcome and signs it off; any number of roles can be Responsible for doing the work. Use A/R where the accountable role also does the work.
Who should be accountable for AI security overall?
In the sample the CISO is accountable for most security activities, the AI product owner for decisions about their own system (agent permissions, testing authorisation, release), and the executive sponsor for scope, policy and residual risk. Change it to match how your organisation actually decides.
Is this a finished document I can rely on?
No. Every tool in this suite is a template with sample content. It gives you a structured, editable starting point and generates a document from your own answers, but it is not legal, audit, certification, investment or professional advice. Review and adapt the output, and have it checked by the appropriate specialist before you rely on it.
Who can use it?
The page and sample are public. A free account gets 3 free tool runs, shared across every tool; after that, customising the template with your own details and exporting the generated document (Word, PDF, Markdown, CSV) needs Pro, $5/month or $40/year. Your work is saved in your own browser.
ISO 27090 toolkit
Explore all AI & ISO standards →Everything on PlayCISO for ISO/IEC 27090, from the first self-check to the auditor’s questions. ISO/IEC 27090 is guidance, not a certifiable standard.
ISO 27090 Readiness Checklist
Reference · GuideCan you be certified to ISO 27090?
Planner · ProISO 27090 Project Plan
Roadmap · ProISO 27090 Maturity Timeline
Templates · ProISO 27090 Sample Artefacts
Lab · ProISO 27090 Implementation Lab
Lab · ProISO 27090 Threat-Mapping Lab
Template · ProISO 27090 AI Security Test Plan
Template · ProISO 27090 Project Plan Template
Matrix · FreeISO 27090 Traceability Matrix
Generator · FreeISO 27090 Alignment Statement
Labs · FreeISO 27090 Practice Path
Matrix · FreeAI Shared-Responsibility Matrix
Briefing · FreeAI Security Board Briefing Pack
Metrics · FreeAI Security Metrics Set
Estimator · FreeISO 27090 Cost Estimator
Calendar · FreeAI Security Review Calendar
Course · FreeISO 27090 in 10 Minutes
Simulator · FreeISO 27090 Auditor Interview
Article · GuideISO 27090 explained
ISO 42001 and related tools
AI standards toolkit →Every ISO tool on PlayCISO, grouped by the standard it supports. Standards are described in our own words; you still need your own copies of them.
ISO/IEC 42001: AI management system
The certifiable AI management system standard: clauses 4 to 10 plus Annex A controls, audited by an accredited certification body.
- ISO 42001 AIMS Scope BuilderFreeClause 4: context, interested parties and a draft scope.
- ISO 42001 Gap Assessment LabProPractise judging conformity against the AIMS standard.
- ISO 42001 RACI TemplateProTen roles × clauses 4–10 and Annex A areas A.2–A.10.
- ISO 42001 Implementation Project PlanProScope to Stage 2: phases, tasks, owners and indicative durations.
- ISO 42001 Annex A GuideFreeA.2 to A.10: what good looks like and audit evidence.
- ISO 42001 SoA BuilderProStatement of Applicability with status and justification.
- ISO 42001 Objectives PlannerFreeMeasurable AI objectives with owners and KPIs.
- ISO 42001 AI Competence MatrixFreeClauses 7.2 and 7.3: roles, target levels and gaps.
- ISO 42001 Documented Information RegisterFreeEvery AIMS document and record with owner and retention.
- ISO 42001 Evidence BinderProAudit evidence by clause and Annex A area.
- ISO 42001 Internal Audit PlannerFree9.2: a risk-based 12-month audit programme.
- ISO 42001 Management Review PackPro9.3: every input, KPIs, decisions and actions.
- ISO 42001 Nonconformity TrackerFree10.2: root cause, corrective action, effectiveness.
- ISO 42001 Mock AuditProRehearse the certification audit and get a debrief.
- ISO 42001 Stage 2 Audit LabFreeThree audit days with a certification auditor.
- ISO 42001 Certification CountdownFreePlan back from your certificate date.
- ISO 42001 + 27001 IMS PlannerFreeWhat an AIMS and an ISMS can share or merge.
ISO/IEC 42005: AI system impact assessment
How to assess an AI system's impact on people, groups and society; feeds the 42001 impact assessment.
- ISO 42005 AI Impact AssessmentProAssess impact on people, groups and society.
ISO/IEC 23894: AI risk management
Guidance on managing AI risk, built on ISO 31000; feeds the 42001 risk assessment and treatment.
- ISO 23894 AI Risk Treatment PlanProIdentify, analyse and treat AI risks with owners.
Across standards
Implement a control once and evidence it for every framework that asks for it.
- AI Standards CrosswalkPro42001, 27001, 27090, EU AI Act and NIST AI RMF side by side.
More security tools
The tools people use most on PlayCISO.
Catch injected instructions hiding in content your AI agent should only read.
Score your readiness against the first international standard for AI security threats.
Critical CVEs across 21 major vendors, month by month — and what the 2026 surge means.
Build an ordered ruleset, replay labelled packets and catch shadowed and any-to-any rules.
Self-assess all six Functions, from Govern to Recover, and see your weakest one.
Search 172 threat actors across Microsoft, CrowdStrike, Mandiant and community names.