Role
Third-party risk manager
Assuring somebody else’s system, with the duty still sitting with you.
§ 1 — What you carry
Where the duty attaches
Almost nothing in this corpus lets an organisation transfer a duty to its supplier. What the instruments provide for is a duty to obtain and keep evidence about the supplier — which is a different thing, and it is why a completed questionnaire discharges very little on its own.
The distinct capacity worth knowing is the external assurance provider: a firm reporting to a third party is neither the internal audit function nor a certification body, and the regime that binds it does so precisely because it is external to the entity it reports on.
Also written as vendor risk manager, supplier assurance lead, outsourcing manager. Everything below is drawn from the obligation dataset by the capacity that binds this role, not by job title — 11 provisions across 2 collections.
§ 2 — The obligations
What binds you
Agentic AI oversight
6 provisions
- IMDA Agentic AI Framework v1.5 §2.1 — Assess and bound the risks upfront
- IMDA Agentic AI Framework v1.5 §2.1.1 and §2.2.2 — Reversibility of agent’s actions
- IMDA Agentic AI Framework v1.5 §2.3.3 — Continuous testing and monitoring
- MAS Consultation Paper P017-2025 — Consultation Paper on Guidelines on Artificial Intelligence Risk Management
- Model AI Governance Framework for Agentic AI v1.5, §§2.1.2 and 2.3.1 — Bound risks through design by defining agents limits and permissions
- Model AI Governance Framework for Agentic AI v1.5, §2.3.2 — Before deploying, test agents
EU AI Act
5 provisions
- Article 26 — Obligations of deployers of high-risk AI systems
- Article 27 — Fundamental rights impact assessment for high-risk AI systems
- Article 5 — Prohibited AI practices
- Article 50 — Transparency obligations for providers and deployers of certain AI systems
- Article 6 — Classification rules for high-risk AI systems
§ 3 — Evidence
The artefacts this role owns
Drawn from the evidence column of the provisions above, most widely demanded first. These are what an assessor asks this role to produce.
- 1xA correlation identifier surviving hand-offs
- 1xA dated classification determination per system
- 1xA dated residual-risk acceptance with a named accepter
- 1xA handling position for trajectories carrying personal data
- 1xA justification line against every write permission
- 1xA log entry showing a runtime constraint firing
- 1xA log-layer coverage matrix
- 1xA named oversight assignment, with the competence behind it
- 1xA negative-control suite, with results
- 1xA permission matrix at tool, verb and object granularity
- 1xA privilege-drift reconciliation, approved against running
- 1xA re-review trigger tied to change
§ 4 — Ahead
What lands next
2 December 2026
2 December 2026
Article 50 — Transparency obligations for providers and deployers of certain AI systems
2 December 2027
Article 26 — Obligations of deployers of high-risk AI systems
2 December 2027
Article 27 — Fundamental rights impact assessment for high-risk AI systems
2 December 2027
2 August 2028
Article 26 — Obligations of deployers of high-risk AI systems
2 August 2028
Certification
Assessed on the same standard of evidence
Every Council credential is examined on applied judgement against a published anchor, set out the way the obligations on this page are. The Certified Applied AI Associate, the Council’s flagship, is open to any adult today, its fee waived by a coupon code from the AI-Readiness Check, and the register lists what is open for enrolment.