Artificial intelligence
AI risk management asks what a system can do, how it fails, who is affected and who is accountable, across a lifecycle where capability can change after deployment and harms can be diffuse.
- 01In briefThe short answer
- 02Three mandatesThe field, seen three ways
- 03The systemic questionFrom model failure to systemic effect
- 04FrameworksRisk frameworks and law
- 05CouplingsWhere AI risk lands
- 06GroundingThe methods behind the field
- 07Key termsTerms used on this page
GCRI · Institution · service level R0
The short answer.
Artificial intelligence as a field of risk covers how AI systems are built, evaluated, deployed and governed, and the safety, security, rights and societal effects that follow from their use. In the Nexus Consortium it is studied as science at GCRI by Artificial Intelligence and Autonomous Systems as well as Future of Work and Skills, governed at GRF by Innovation Nexus as well as AI and Digital Governance, and understood financially at GRA by AI and Financial Model Risk; this gateway links those platforms and the methods they share.
Digital demand and connectivity in East Asia.
Urban demand centres and the hubs that anchor digital and data infrastructure, shown for East Asia as an example geography for this subject.
The field, seen three ways.
From model failure to systemic effect.
Individual AI failures (biased outputs, hallucinated facts, unsafe actions by an agent) are well catalogued. Systemic effects arise when many actors use similar models trained on similar data, when AI is embedded in critical decisions such as credit or triage, or when capability shifts faster than oversight. Assessment has to cover the model, the system it sits in and the institution operating it.
| Layer | Typical risk | Assessment |
|---|---|---|
| Model | Robustness, bias, capability for misuse | Evaluations and red teaming |
| System | Integration errors, over-reliance, data leakage | Impact assessment, human oversight design |
| Institution | Unclear accountability, missing incident response | Management system audit (ISO/IEC 42001) |
| Ecosystem | Homogeneity, concentration, correlated failure | Market and dependency analysis |
Risk frameworks and law.
| Framework or body | What it governs or supplies |
|---|---|
| NIST AI RMF and its generative AI profile | Govern, map, measure and manage functions |
| ISO/IEC 42001 and ISO/IEC 23894 | AI management systems and AI risk management guidance |
| EU AI Act | Risk-tiered obligations applied in stages; check current dates for each obligation |
| OECD AI Principles and UNESCO Recommendation on the Ethics of AI | Intergovernmental principles |
Where AI risk lands.
- Compute: capability concentration and energy demand.
- Work: task exposure and transition.
- Information: synthetic content and trust.
- Financial stability: model risk and herding.
The science sits on AI and Autonomous Systems; methods start from AI assurance.
The methods behind the field.
- Artificial intelligence and machine learning
- AI evaluation and assurance engineering
- High-performance, cloud and edge computing
- Probability and statistics
- Cybersecurity and cryptography
- Social and behavioural science
- Robotics and automation
- Economics and development economics
- Econometrics
- Policy, institutional and decision science
- Systems engineering
- Digital identity and public digital systems
- Uncertainty quantification and model validation
Terms used on this page.
Each term as this page uses it. Where a standard-setting body defines the term, the wording follows that body.
| Term | Definition |
|---|---|
| AI system | A machine-based system that infers from inputs how to generate outputs such as predictions, content or decisions that can influence environments, in OECD usage. |
| Deployer | In the EU AI Act, a person or organization using an AI system under its authority, distinct from the provider that develops it. |
| Model risk | The potential for adverse consequences from decisions based on incorrect or misused model outputs, as SR 11-7 describes it. |
| General-purpose AI model | An AI model that can perform a wide range of distinct tasks and be integrated into many applications. |
Where this page's references lead.
The bodies and documents this page is read against. Status and editions change; each link goes to the issuing body.
Questions and answers.
Why does GCRI host field gateways?
Around GCRI.
Start with an enquiry.
The current service level is R0: information and enquiries. An enquiry receives a reference and names the next step; applications, payments, compute allocations and registrations open only when their team, terms and review capacity exist.