Standaarde vir risiko van KI en outonome stelsels
AI and autonomous systems are being deployed into decisions about credit, health, infrastructure and security faster than their failure modes are understood, and general-purpose models create common dependencies across sectors. This area of the Digital, Cyber and AI programme sets records for model documentation, evaluations, incidents and deployment contexts, crosswalked to AI governance frameworks, so that risk claims about AI systems rest on documented evidence. Each artefact of the area carries a UNI (Unique Nexus Identifier).
- Nexus Standards: NS-0, NS-10, NS-11, NS-12.
- Open standards: AI Risk Management Framework (AI RMF 1.0); OECD AI Principles; Regulation (EU) 2024/1689 (Artificial Intelligence Act); PROV-O: The PROV Ontology; Verifiable Credentials Data Model v2.0.
- Data and feeds: OSV open source vulnerability database; Our World in Data datasets.
- Couples with: Cyber and digital; Information integrity and trust; Frontier technology; Financial stability and prudential risk.
Summary
- 01ScopeScope and boundaries
- 02Hazard cascadesCascading and compounding hazard pathways
- 03Hazards and sectorsHazards, threats and sectors concerned
- 04System modelHow a shock travels through this system
- 05Data sourcesData sources
- 06Sistemiese koppelingsSistemiese koppelings
- 07Open standardsOpen standards referenced
- 08Restricted standardsRestricted compatibility profiles
- Nexus Standards
- 4
- NS-0, NS-10, NS-11
- Open standards
- 5
- named, never copied
- Data and feeds
- 2
- listed for reference
- Couplings
- 4
- with other areas
- UNI
- requested · allocation in progress
- Status
- Proposedin the catalogue · planned · scope LAB
- Last reviewed
- area profile
Nexus Standards (2026). AI and autonomous systems risk standards: area brief, Digital, Cyber and AI programme. Proposed. https://nexustandards.com/digital-ai/ai-and-autonomous-systems-risk-standards/. Last reviewed 17 October 2026.
Files include the credits and the licence statement and are generated in the browser; no data is transmitted. Downloads require JavaScript.
Scope and boundaries
An area of the Digital, Cyber and AI programme (hazard and threat families). The area profile sets out 5 matters in scope and 3 matters addressed by other areas.
In scope 5
- Model and dataset documentation records (model cards, datasheets)
- Evaluation and red-team result records with method
- AI incident records
- Deployment context and risk-tier records (NS-9)
- Machine-readable AI obligations (NS-10)
Out of scope 3
- Model training techniques
- Release of capability-enhancing details for dangerous uses
- Product endorsements
Jurisdictions Application in each jurisdiction is a national decision, taken by the National Standards Council (NSC) with the competent national authorities.
Cascading and compounding hazard pathways
Each hazard named for this context, as classified in the Nexus Hazard Standard (NHS-1), with the hazards it triggers, amplifies or compounds, the confidence of each relation, the physical or social mechanism, and the next step in the cascade.
The interactive view requires JavaScript; the sections below present the same records.
From detection to evidence
An observation from OSV open source vulnerability database becomes an identified hazard event; a forecast and a warning cite the same event and version; a trigger written as a clause starts anticipatory action for model failure and bias; response and losses are recorded; evidence reaches SDG 9.5.1 and the areas coupled to this one. 9 operations in 5 phases, built from the area profile: the rows carry its own data sources, indicators, open standards, couplings and, where listed, hazards, and each operation names the rules it implements (13). Records and artefacts are cited by UNI (Unique Nexus Identifier) and version.
Prosesverkenner · Standards · kodifiseer en korrigeer
Lewensiklus van 'n standaard
17 stappe·6 fases
Standards · kodifiseer en korrigeer
Lewensiklus van 'n standaard
Lewensiklus van 'n standaard · NS-13-spoor · Bydraekanale · IETF-spoor · W3C-spoor · OGC-spoor · OASIS-spoor · ISO- en IEC-fases · Konformiteit · Rekordlewensiklus · Roetering · Monitering.
Siklusbevindings word veranderingsvoorstelle, die Wêreldwye Standaarderaad aanvaar ná 'n openbare laaste oproep, en elke land besluit oor sy eie aanvaarding.
01 Lei sendings
- 01Route missions to standards and profilesStep · Nexus StandardsNexus Standards routes each candidate mission to the standards and profiles that apply.
02 Profiele en ooreenstemming
- 02Stel sendingprofiele saamStap · Nexus StandardsNexus Standards stel sendingprofiele saam en teken enige nasionale aanvaarding aan waarvan die sendings afhang.
- 03Valideer bou-uitsette teen die profieleStap · Nexus StandardsNexus Standards valideer bou-uitsette teen die saamgestelde profiele en teken die bewyse aan.
- 04Moniteer konformiteit en teken verseëlde bewyse aanStap · Nexus StandardsGedurende die Aksieweek monitor Nexus Standards konformiteit en teken afwykings aan.
03 Voorstelle en beoordeling
- 05Afwykings en regstellings word veranderingsvoorstelleStap · Nexus StandardsAfwykings word veranderingsvoorstelle, en die regstelling word as 'n verandering aan die bewyse aangeteken.
- 06Siklusbevindings word veranderingsvoorstelleStap · Nexus StandardsBevindings uit 'n siklus word veranderingsvoorstelle: metode- en korreksiebevindings uit Verslae (F06) en dokketitems uit Nexus…
- 07Stel op en hersien in die openbaarStap · Komitees en werkgroepeTegniese komitees en werkgroepe stel die standaarde op en hersien dit in die oop proses van NS-13.
- 08Beoordeel met die oog op 'n raadsbesluitStap · Nexus StandardsNexus Standards beweeg veranderingsvoorstelle deur beoordeling met die oog op 'n raadsbesluit.
04 Raadsbesluit
- 09Openbare laaste oproepStap · Global Standards CouncilDie Global Standards Council aanvaar by 'n tweederdemeerderheid ná 'n laaste oproep van negentig dae.
- 10Wêreldwye aanvaarding van 'n standaardBesluit · Global Standards CouncilDie Wêreldwye Standaarderaad besluit oor die aanvaarding van elke standaard ná 'n openbare laaste oproep.
05 Publiseer en identifiseer
- 11Assign a UNI and publishStep · Secretariat (GCRI Canada)Nexus Standards assigns UNIs and deposits publications under the publisher’s process.
- 12'n UNI identifiseer en sertifiseer nooit nieRekord · Nexus StandardsElke objek van die programjaar dra een identifiseerder, 'n UNI onder NS-1, sodat dit aangehaal, geweergaweer en opgelos kan word.
- 13Publiseer die jaar se aanvaarde veranderingeStap · Nexus StandardsBy hernuwing publiseer Nexus Standards die jaar se aanvaarde veranderinge en die nasionale aanvaardingstatus.
06 Nasionale aanvaarding en toepassing
- 14Nasionale aanvaarding deur die Nasionale StandaarderaadBesluit · Nasionale StandaarderaadElke land se Nasionale Standaarderaad besluit oor nasionale aanvaarding: aanvaar soos dit is, aanvaar met 'n profiel (NS-16), pas aan, of verwys slegs.
- 15Aangenome standaarde bereik kursusse en uitgawesStap · Nexus StandardsElke aangenome standaard of profiel word 'n kursusmodule by Risk Academy (F09), en standaarde word in Reports-uitgawes (R1) toegepas.
- 16Aanvaarding is finaal vir elke jurisdiksieStandaard · Nasionale StandaarderaadAanvaarding is finaal vir elke jurisdiksie, en die GSC het geen veto oor 'n nasionale besluit nie. Geen globale liggaam hou 'n veto nie.
- 17Nasionale aanvaardingsrekordsRekord · Nexus StandardsSaamgestelde missieprofiele en nasionale aanvaardingsrekords gaan na die Foundry by gereedheidsvoorbereiding (F08).
Nexus Standards · Nexus Standards (institution) · National Standards CouncilAn outline of the designed programme cycle, not a record of operation: no cycle has run and every programme is at service level R0. Every Nexus Standard is at the PROPOSED stage and nexustandards.com governs the text and status of each one. A UNI identifies and never certifies; a validator report is evidence for review, not a conformance claim.
The operations, as a table
| Step | Phase | Operation | Who | Decision rule | Record | Implements |
|---|---|---|---|---|---|---|
| 01 | Detection | Observe (start). The observation is recorded with its station or sensor metadata and source release. Data sources of this area: OSV open source vulnerability database and Our World in Data datasets. | the provider of the observation | each observation is a record with its source release and version | observation records from OSV open source vulnerability database, each with a record UNI (NS-3) | NS1-R2, NS2-R1 |
| 02 | Detection | Hazard event (record). The event receives an identifier that forecasts, warnings, triggers and losses cite, so that each can be traced to the same event. Hazards in scope: Model failure and bias, Misuse of general-purpose AI, Loss of human oversight, AI-enabled cyber attacks and Correlated failure from shared models. | the competent hazard service | one event identifier, cited by every forecast, warning and loss record | a hazard event record for model failure and bias, classified by the hazard information profiles | NS1-R1 |
| 03 | Forecast and warning | Forecast (model). Ensemble forecasts carry lead time and verification statistics, which determine how reliable a trigger can be. | the forecasting centre | a forecast cites the event and its own version; verification statistics are recorded | a forecast record from Our World in Data datasets with lead time and version | NS2-R2, NS12-R2 |
| 04 | Forecast and warning | Warning (record). The warning is issued by the authority designated under national law and carries the identifiers of the event and forecast it rests on. | the designated alerting authority | the warning cites the event identifier and forecast version | an impact-based warning in the Common Alerting Protocol | NS2-R1 |
| 05 | Anticipatory action | Trigger clause (gate). The threshold, lead time and forecast source are stated in advance; an evaluation with missing inputs is indeterminate, never a trigger. | the anticipatory action partners, under a pre-agreed protocol | the trigger is a clause whose predicate is data; two evaluators reach the same result | a trigger written as a clause (NS-10) for model failure and bias | NS10-R1, NS10-R3, NS11-R3 |
| 06 | Anticipatory action | Early action (step). Actions agreed in advance are taken before the peak impact. Sectors concerned: Digital, telecom and platforms, Finance, insurance and capital markets, Health systems and Public sector and government. | public agencies and humanitarian partners | funds and actions are released on the attested trigger | pre-agreed actions released, with the attestation of the trigger | NS11-R1 |
| 07 | Reaksie | Response and loss (record). Damage and loss are recorded against the event, so that warnings, actions and impacts can be compared. | the national disaster management agency | losses cite the event identifier | impact and loss records disaggregated as reporting requires | NS2-R4 |
| 08 | Bewyse | Report 9.5.1 (end). Reporting draws on the recorded events, warnings and losses. Indicators of this area: SDG 9.5.1 (Research and development expenditure as a proportion of GDP). | the national reporting focal point | the indicator is computed from records and cites them | SDG 9.5.1 (Research and development expenditure as a proportion of GDP) | NS11-R5, NS14-R1 |
| 09 | Bewyse | Coupled areas (end). Common driver: AI systems introduce new attack surfaces and are used to scale attacks. Coupled areas: Cyber and digital risk standards, Information integrity and trust, Frontier technology risk standards and Financial stability and prudential risk. | the programmes of the coupled areas | shared events are cited, never copied | records shared with Cyber and digital risk standards | NS0-P9 |
Table.Operations in walking order; each row names the phase, the operation, its actor, its decision rule, the record it leaves and the rule identifiers it implements. The table shows the process as specified; it does not report the status of any standard, proposal or record. Sources: AI Risk Management Framework (AI RMF 1.0) · OECD AI Principles · NS-0: Constitution and principles · NS-10: Policies and clauses.
Hazards, threats and sectors concerned
Hazards and threats 5
Sektore 4
How a shock travels through this system
A system model of this area built from the registry: 11 hazards, 5 systems, 4 sectors, 10 controls, 2 feeds, 1 indicator. Select the components to include, choose an initiating hazard and run the scenario. Stress propagates through the hazard relations of the Nexus Hazard Standard (NHS-1), the exposure of each system, the recorded couplings and the sectors that carry the loss; adopted standards dampen transmission and live data sources support detection. All parameters are stated assumptions that the reader can adjust; results are illustrative scenario outputs for exercises and training, not forecasts or risk assessments. The Map view places the same run on an example geography, North America, the region of the programme’s context map, with open data: country boundaries, a grid of hexagonal cells and NASA satellite layers for the initiating hazard. The geography serves exercises and training and is not an assessment of that place.
Gevare
- Dam failure
- Sea-ice loss
- Coastal erosion (set off downstream)
- River flood (set off downstream)
- Flash flood (set off downstream)
- Mass displacement (set off downstream)
- Biodiversity loss (set off downstream)
- Widespread power outage (set off downstream)
- Water supply shortfall (set off downstream)
- Arctic winter sea-ice collapse (set off downstream)
- Polar vortex disruption (set off downstream)
Systems and couplings
Controls
NS-0Grondwet en beginselsNS-10Beleide en klousules (masjienleesbare verpligtinge: definisie, binding, evaluering en afdwinging)NS-11Verifikasie en monitering (getuienis, attestasies, zero-trust-ooreenstemmingwaarneming)NS-12Sagteware-, model- en KI/ML-artefakte (voorsieningsketting, modelkaarte, datastelblaaie)NRS-H100Gevarefamilie: Robotika, outonomie en outonome wapens (NHS-1 tech.autonomy-weapons)NRS-H53Gevarefamilie: Kuber-fisiese ontwrigting (NHS-1 cyber.cyber-physical)NRS-H61Gevarefamilie: Kaskaderende faling (NHS-1 sys.cascade)NRS-H113Gevarefamilie: Broosheid van die finansiële stelsel (NHS-1 soc.financial-system)NRS-A8KI-gereed risikodata en modelverpakkingNRS-H54Gevarefamilie: Sein- en platformontwrigting (NHS-1 cyber.platform)
Observation
- OSV open source vulnerability database (listed for reference)
- Our World in Data datasets (listed for reference)
- Research and development expenditure as a proportion of GDP
Example geography
- North America: 243 hexagonal cells
- Night lights tonight, outages (VIIRS Day/Night Band), NASA GIBS
- Surface water through cloud, 30 m (OPERA DSWx-S1 radar), NASA GIBS
- Rain rate, last half hour (IMERG), NASA GIBS
- Vegetation disturbance alerts, 30 m (OPERA DIST-ALERT), NASA GIBS
Data sources
2 data sources named in the area profile, each with provider, licence, update frequency and delivery path.
Data sources are listed for reference; live data is displayed once licence review is complete.
- Lewendig
- 0
- on completion of licence review
- Listed in the data register
- 0
- licence terms recorded
- Proposed for the register
- 2
- licence terms to be recorded
- Delivery
- 0 governed service · 0 plugin · 0 browser · 1 snapshot · 1 listed for reference
- proposed paths in order of preference: governed data service, plugin, direct browser read, build-time snapshot; sources whose terms do not permit reuse are listed for reference only
OSV open source vulnerability databasefor reference
OSV project (Google, OpenSSF) · continuous
Per-source licences; OSV schema CC BY 4.0
listed for referenceproposed for the data register.
Our World in Data datasetsfor reference
Global Change Data Lab (Our World in Data) · varies.
CC BY 4.0 for OWID-produced data; third-party data keep their licences.
build-time snapshotproposed for the data register
Sistemiese koppelings
4 couplings link this area to other areas and programmes, each with the mechanism of the link.
- Cyber and digital risk standardsDigital, Cyber and AICommon driver: AI systems introduce new attack surfaces and are used to scale attacks.
- Information integrity and trustGovernance, Security and SocietyCascade: generative AI lowers the cost of synthetic disinformation.
- Frontier technology risk standardsDigital, Cyber and AIFeedback: AI accelerates other frontier technologies, including biotechnology.
- Financial stability and prudential riskFinance, Insurance and the EconomyCommon driver: shared models and data can produce herding and correlated failures.
Open standards referenced
5 open standards, each with its relation as recorded in the register under NS-14. Hover over or focus a relation for its definition.
adopt2
PROV-O: The PROV OntologyW3Clisted in the register · w3c-prov-oVerifiable Credentials Data Model v2.0W3Clisted in the register · w3c-verifiable-credentials-2Restricted compatibility profiles
2 standards whose text is not openly available, named and linked only. Any compatibility profile is kept on protected pages until an arrangement with the issuing body is in place.
- ISO/IEC 42001:2023 Artificial intelligence – Management systemISO/IEC JTC 1/SC 42compatibility profile (restricted)
- ISO/IEC 23894:2023 Artificial intelligence – Guidance on risk managementISO/IEC JTC 1/SC 42compatibility profile (restricted)
Aanwysers
The area profile names 1 indicator, each with its framework and code.
Indicator names and codes are given as published by their stewards.
Standards that apply
The area profile lists 4 Nexus Standards; the registry lists NS-0, NS-3, NS-9, NS-11, NS-12 for this area. Each is a proposed standard under development.
- 01
NS-0 · Constitution and principles
Status: proposed · 12 requirements stated. Read NS-0 · Requirements for this area.
- 02
NS-10 · Policies and clauses (machine-readable obligations: definition, binding, evaluation and enforcement)
Status: proposed · 7 requirements stated. Read NS-10 · Requirements for this area.
- 03
NS-11 · Verification and monitoring (evidence, attestations, zero-trust conformance observation)
Status: proposed · 6 requirements stated. Read NS-11 · Requirements for this area.
- 04
NS-12 · Software, model and AI/ML artefacts (supply chain, model cards, dataset sheets)
Status: proposed · 4 requirements stated. Read NS-12 · Requirements for this area.
Routing, monitoring and verification tools
Verification console
Test suites, conformance statements and attestations for each standard, as recorded in the registry.
PlannedDeployment
Reference deployments and a sandbox with worked examples.
In progressMonitoring console
Version and source status for each standard, as published by the issuing bodies.
Each console computes from the registry and states its reasoning; decisions remain with the competent bodies.
Working group charter
A proposed charter for this area, open to institutions and experts.
Model, evaluation, incident and deployment-context records crosswalked to AI governance frameworks.
Institutions and experts are invited to charter this working group under NS-13 by filing a public change record. National Working Groups (NWGs) contribute national inputs and field exercises through their National Nexus Consortium (NNC).
- AI evaluation record profile
- AI incident record schema
- Obligation crosswalk to AI RMF functions and EU AI Act risk tiers (links only)
- NIST AI Safety Institute Consortium liaison sought
- OECD.AI Network of Experts liaison sought
- MLCommons liaison sought
- Partnership on AI liaison sought
Evidence, recognition and capital pathways
What each institution of the ecosystem draws from this area: The Global Centre for Risk and Innovation (GCRI Canada), as Secretariat, for evidence; The Global Risks Forum (GRF) for recognition; and The Global Risks Alliance (GRA) for capital.
GCRI Canada
GCRI Canada may accept evaluation and incident records with method, version and evaluator as research evidence; the decision rests with GCRI Canada under its own evidence and research rules.
Die Global Risks Forum (GRF)
GRF may record organisations publishing model documentation on the shared profile as a dialogue item; the decision rests with GRF under its own recognition rules.
Die Global Risks Alliance (GRA)
GRA may read AI risk-tier and evaluation records when it considers AI-dependent investments; the decision rests with GRA under its own capital-readiness rules.
Each institution determines under its own rules what it accepts as evidence; a pathway reading is neither an approval nor investment advice.
Open research questions
3 questions identified as priorities for the working group.
- Evaluations of general-purpose models are not standardised and can be gamed.
- There is no shared global AI incident taxonomy.
- Systemic risk from model concentration is not measured.
Intended users
The area profile identifies 4 user communities.
- AI developers and evaluators
- AI governance and assurance teams
- Regulators
- AI safety researchers
Registry listing and depth pages
This area is in the catalogue (status: planned; scope LAB; 4 crosswalk domains). It has five depth pages and a register crosswalk of its domains.
Standards that apply
The rules of each listed standard, read through the specification reader.
Depth · CrosswalkKruisverwysing
Every register entry in the area’s domains, by relation, with the crosswalk graph.
Depth · Records and schemasRecords and schemas
Scope, record, schema explorer and how a UNI resolves.
Depth · ConformanceKonformiteit
Clauses, evidence and attestations for conformance assessment.
Depth · ReadinessGereedheid
The area alongside the three pathways, as a matrix.
The crosswalk as a table: 47 external standards by domain
Nexus Standards in the graph: NS-0, NS-3, NS-9, NS-11, NS-12. A relation is a register entry, never a compliance claim.
5 Nexus Standards · 4 domains · 47 external standards · 23 issuing bodies · 0 implementing repositories. Relations (adopt, profile, crosswalk, cite) are as recorded in the register; names and standards remain with their stewards.
Crosswalk graph79 nodes and 114 links from the registry of external standards, programme areas and implementations (the external standards register, the consumers register. Relations (adopt, profile, crosswalk, cite) are shown as recorded in the register.
Areas carry no geography of their own; the map shows where the bodies that issue the area’s external standards are based.
Issuing bodies and their seats (16)
| Body | Seat | Register entries |
|---|---|---|
| NIST | Gaithersburg, Maryland, United States | 10 |
| ISO/IEC JTC 1/SC 42 | Geneva, Switzerland | 6 |
| European Parliament and Council | Brussels, Belgium | 4 |
| ISO/IEC JTC 1/SC 27 | Geneva, Switzerland | 4 |
| IEEE Standards Association | Piscataway, New Jersey, United States | 3 |
| OASIS Cyber Threat Intelligence TC | Woburn, Massachusetts, United States | 2 |
| Agentic AI Foundation (Linux Foundation) | San Francisco, California, United States | 1 |
| CEN-CENELEC JTC 21 | Brussels, Belgium | 1 |
| FIRST CVSS Special Interest Group | Cary, North Carolina, United States | 1 |
| IETF | Wilmington, Delaware, United States | 1 |
| ISO/IEC JTC 1/SC 7 | Geneva, Switzerland | 1 |
| ISO/TC 292 | Geneva, Switzerland | 1 |
| OASIS CSAF TC | Woburn, Massachusetts, United States | 1 |
| OpenSSF (Linux Foundation) | San Francisco, California, United States | 1 |
| OpenVEX community (OpenSSF project) | San Francisco, California, United States | 1 |
| UN Office of the Secretary-General’s Envoy on Technology and UNDP | New York, New York, United States | 1 |
Issuing bodies are shown at the city of their public seat. The 1 programmes, 0 implementing repositories and 19 Nexus Standards have no location and are not mapped · 7 issuing bodies have no single public seat and are listed without a location: Mitchell et al. (FAT* 2019); World Customs Organization; OECD; UNESCO; United Nations; International Atomic Energy Agency; and others.
Standards globe16 issuing bodies with a public seat and 47 register entries, shown in equal-area cells at four levels of detail. Seats are placed at city coordinates; programme areas, implementations and standards have no location and are not drawn. A still world map is shown where scripts are unavailable.
Bronne
4 references from peer-reviewed and official publications cited in the area profile, linked at the publishers’ own sites.
peer-reviewedManaging extreme AI risks amid rapid progressScience · 2024peer-reviewedDatasheets for datasetsCommunications of the ACM · 2021peer-reviewedModel Cards for Model ReportingProceedings of the Conference on Fairness, Accountability, and Transparency (ACM) · 2019- officialInternational AI Safety Report 2025UK Department for Science, Innovation and Technology (for the report’s expert advisory panel) · 2025
Recent research on this area
Recent peer-reviewed articles, preprints, datasets and reports on the subject of this area, retrieved at the time of reading from open scholarly metadata services (OpenAlex, Crossref, DataCite, Zenodo, arXiv, Europe PMC, OpenAIRE Graph) with search terms composed from the registry. Each work is listed with its authors, venue, year, type, open-access status and licence, and links to its DOI or landing page. Sources not yet enabled for live display on this site are listed for reference and are not queried.
Search termsAI safety; AI risk; AI risks; AI incidents; AI harms; frontier AI; general-purpose AI; AI misuse; AI accidents; autonomous weapons; AI Risk Management Framework; OECD AI PrinciplesDerived from the title of the area and the hazards in its profile, the NHS-1 hazards it names and the open standards it lists.
- OpenAlexlisted for reference; not queried
- Crossreflisted for reference; not queried
- DataCitelisted for reference; not queried
- Zenodolisted for reference; not queried
arXivlisted for reference; not queried- Europe PMClisted for reference; not queried
- OpenAIRE Graphlisted for reference; not queried
No source is enabled for live display on this site yet: the sources are listed for reference and none is queried. Once the publisher of record enables a source after its terms are reviewed, recent works are listed here. Metadata and links only: title, authors, venue, year, type, open-access status and licence as each source records them; works are listed for orientation, not reviewed or endorsed.
Risk Academy courses and disciplines
2 courses and 4 disciplines of Risk Academy (risksacademy.com) that teach the knowledge and methods of this area. Each course lists its level, stream and learning hours.
Kursusse
Dissiplines
The correspondence between courses, disciplines and the Nexus Standards is maintained with Risk Academy and is subject to its academic review.
Programmes, tracks and courses for this area
GCRI Canada, The Global Risks Forum (GRF), The Global Risks Alliance (GRA) and Risk Academy work on the subject of this area in distinct roles (research and evidence; governance, policy and convening; capital, finance and markets; education and training). Each entry states why it corresponds and the step it offers; each institution sets its own terms of participation.
Navorsing en bewyse
GCRI Canada
Research programmes that produce the observations, models and assessments these records describe.
Artificial Intelligence and Autonomous SystemsStudies model evaluation, agentic systems and autonomous physical systems.Apply to this programme therisk.globalAll programmes of GCRI CanadaGovernance, policy and convening
Die Global Risks Forum (GRF)
Tracks that convene public authorities, researchers and institutions on the governance of shared risks.
AI and Digital GovernanceExamines AI oversight and accountability, and liability for autonomous agents.Join this track globalriskforum.comAll tracks of The Global Risks ForumCapital, finance and markets
Die Global Risks Alliance (GRA)
Industry programmes in which financial institutions build a shared understanding of systemic risk.
AI and Financial Model RiskWorks on model risk governance and generative AI in finance.Join this programme globalriskalliance.comAll programmes of The Global Risks AllianceOnderwys en opleiding
Risk Academy
Courses and disciplines that teach the knowledge and methods practitioners need to apply these standards.
Technology and DataRAT-204 · Intermediate · 19 learning hoursManaging risk data, AI, digital infrastructure and technology risk responsiblyTake this course Uncertainty and ConfidenceRAS-404 · Intermediate · 29.5 learning hoursCalibrated probability, confidence language and decisions under deep uncertaintyTake this course All courses and disciplinesListed in the education and training section of this pageKernterme
- UNI
- Unique Nexus Identifier: a persistent identifier that resolves an artefact to its record.
- voorgestel
- A draft in the open development process that the Global Standards Council has not yet adopted. Each area profile is a proposal to the Nexus Standards registry.
- crosswalk
- Relation in which clauses of these standards are mapped to parts of the external standard by clause number, without reproducing its text.
- cite
- Relation in which the external standard is cited by identifier and URL as a reference.
- adopt
- Relation in which the external standard is used as published and values are validated against it.
- compatibility profile (restricted)
- Relation to a standard whose text is not openly available: the standard is named and linked only, and any compatibility profile is kept on protected pages.
- change record
- A public record of a proposed change under NS-13, decided by the editors with written reasons.
- attestation
- An evaluator’s recorded result for one clause and one subject under NS-11.
- liaison sought
- A liaison the working group intends to request from the issuing body; a liaison is established only when the body confirms it in writing.
Related areas
Recent research and sources
Peer-reviewed articles, preprints, reports and datasets on the subject of this area, retrieved at the time of reading from open scholarly metadata services (OpenAlex and Crossref) and listed with authors, venue, year, type and access status, with a link to each work. Curated references and related standards are listed first; recent works matched by subject follow and are listed for orientation.
Curated references
- Managing extreme AI risks amid rapid progress
Science · 2024 · Peer-reviewed - International AI Safety Report 2025
UK Department for Science, Innovation and Technology (for the report’s expert advisory panel) · 2025 · Official publication - Model Cards for Model Reporting
Proceedings of the Conference on Fairness, Accountability, and Transparency (ACM) · 2019 · Peer-reviewed - Datasheets for datasets
Communications of the ACM · 2021 · Peer-reviewed
Related standards and frameworks
- AI Risk Management Framework (AI RMF 1.0)
NIST - OECD AI Principles
OECD - Regulation (EU) 2024/1689 (Artificial Intelligence Act)
European Parliament and Council - PROV-O: The PROV Ontology
W3C - Verifiable Credentials Data Model v2.0
W3C
Search terms, derived from the title of the area and the hazards in its profile: AI safety; AI risk; AI risks; AI incidents; AI harms; frontier AI; general-purpose AI; AI misuse; AI accidents; autonomous weapons. With JavaScript enabled, recent works matching these terms are listed from OpenAlex and Crossref.
Questions and answers.
What is the status of the AI and autonomous systems risk standards area?
How are data sources presented?
How can this brief be amended?
Dra bewyse vanaf hierdie bladsy by.
Reports is die Konsortium se kwartaallikse publikasie. Wat jy oor hierdie onderwerp weet, kan bewyse, 'n saak of 'n korreksie in die volgende uitgawe word, aan jou toegeskryf en onder CC BY 4.0 gepubliseer waar dit jou eie werk is.
- E-posPoort GW-STDStuur dit vanaf hierdie bladsyGetuienis, 'n regstelling, 'n geval, 'n metode, 'n beoordeling, 'n vertaling of profielinset. Dit kom aan gemerk met hierdie bladsy, sodat die redakteur wat dit besit dit ontvang.
- ZenodoOpen depositDeposit it and submit to the communityPublish your own record on Zenodo with your ORCID iD, then submit it to the Reports community there. Curators review it; the DOI stays yours.
- ReportsNexus StandardsSee every way to contributeThe contribution types, how each is reviewed, what credit you receive, the licence and what is never accepted.
Stuur nooit persoonlike data, sensitiewe infrastruktuurbesonderhede of enigiets wat u geen reg het om te deel nie. 'n Bydrae word voor gebruik beoordeel; die stuur daarvan verleen geen outeurskap of redaksionele regte nie, en borgskap koop nooit 'n bevinding nie.
Contribute to AI and autonomous systems risk standards
Institutions and experts are invited to propose amendments, join the working group or propose data sources. National participation is organised through each country’s National Nexus Consortium (NNC) and its National Working Groups (NWGs); competent national authorities are engaged through official channels.
Status en gebruik. Die Nexus Standards is voorgestelde standaarde wat deur 'n oop proses onder NS-13 ontwikkel en vir vrywillige gebruik gepubliseer word; hulle maak nie sertifisering, akkreditasie of regulatoriese goedkeuring uit nie, en aanvaarding bly 'n besluit van die Wêreldwye Standaarderaad en, in elke land, van sy Nasionale Standaarderaad. 'n UNI (Unique Nexus Identifier) identifiseer 'n artefak; dit sertifiseer dit nie. Name van standaardliggame en hul standaarde behoort aan hul rentmeesters en word slegs gebruik om die dokumente te identifiseer; geen teks van 'n eksterne standaard word hier gereproduseer nie. Nexus Standards is nie geaffilieer met, onderskryf deur of gesertifiseer deur enige standaardliggaam wat hier genoem word nie.
Lisensies: CC BY 4.0 vir standaardteks, registers, skemas en data; Apache-2.0 vir verwysingskode. Eksterne standaarde word volgens naam, nommer en skakel onder hul eie lisensies aangehaal. Derdeparty-data, -biblioteke en -lettertipes behou hul eie lisensies: sien Kennisgewings van derde partye.