Skip to content

Frameworks

One piece of work, three frameworks

Each of the 34 EU AI Act duties in EU AI Fit is mapped to the ISO/IEC 42001 clauses and NIST AI RMF categories it also supports. In the workspace, every duty shows the same references, so evidence gathered for the Act can be reused for an AI management system or a NIST-based programme.

What this is

A crosswalk: a guide to where the work overlaps. Meeting an EU AI Act duty supports the related clause or category. It does not certify conformity with ISO/IEC 42001 or show that a NIST AI RMF outcome is met.

34 duties

The crosswalk

EU AI Act dutyISO/IEC 42001NIST AI RMF
Maintain context-specific AI literacy measuresArticle 47.2 Competence7.3 AwarenessGOVERN 2 Accountability structures, roles and training
Pause the flagged practice and obtain specialist reviewArticle 56.1.4 AI system impact assessment8.3 AI risk treatmentGOVERN 1 Policies, processes and procedures for AI riskMAP 5 Impacts on individuals, groups and societyMANAGE 1 Prioritising and responding to AI risks
Answer the open classification questionsArticles 5, 6 and 504.1 Understanding the organization and its context6.1.2 AI risk assessmentMAP 1 Context is established and understoodMAP 2 Categorisation of the AI system
Write down the Article 6(3) assessment before the system is suppliedArticle 6(3)–(4)6.1.4 AI system impact assessmentA.5.3 Documentation of AI system impact assessmentsMAP 2 Categorisation of the AI systemMAP 5 Impacts on individuals, groups and society
Register the system treated as not high-risk in the EU databaseArticle 49(2); Annex VIII, Section BA.8.5 Information for interested parties7.5 Documented informationGOVERN 1 Policies, processes and procedures for AI risk
Obtain the provider's Article 6(3) assessmentArticle 6(3)–(4); Article 25A.10.3 SuppliersA.9.4 Intended use of the AI systemGOVERN 6 Third-party software, data and supply chain risksMAP 2 Categorisation of the AI system
Confirm your supplier's AI disclosure and content markingArticle 50(1)–(2) (provider duties; deployer check)A.10.3 SuppliersA.8.2 System documentation and information for usersGOVERN 6 Third-party software, data and supply chain risksMANAGE 3 Third-party AI risks and benefits
Inform people when they interact directly with AIArticle 50(1)A.8.2 System documentation and information for usersA.6.2.2 AI system requirements and specificationMEASURE 2 Evaluation for trustworthy characteristicsMANAGE 2 Maximising benefits and minimising negative impacts
Implement machine-readable marking for synthetic contentArticle 50(2) and (5)A.6.2.2 AI system requirements and specificationA.6.2.4 AI system verification and validationMEASURE 2 Evaluation for trustworthy characteristicsMANAGE 2 Maximising benefits and minimising negative impacts
Notify people exposed to emotion recognition or biometric categorisationArticle 50(3)A.8.5 Information for interested partiesA.5.4 Assessing AI system impact on individuals or groups of individualsMAP 5 Impacts on individuals, groups and societyMEASURE 2 Evaluation for trustworthy characteristics
Label disclosed deepfake contentArticle 50(4)A.9.2 Processes for responsible use of AI systemsA.8.5 Information for interested partiesGOVERN 1 Policies, processes and procedures for AI riskMANAGE 2 Maximising benefits and minimising negative impacts
Disclose AI-generated public-interest textArticle 50(4)A.9.2 Processes for responsible use of AI systemsA.8.5 Information for interested partiesGOVERN 1 Policies, processes and procedures for AI riskMAP 3 Capabilities, intended use, benefits and costs
Establish the high-risk AI risk-management systemArticle 96.1.2 AI risk assessment6.1.3 AI risk treatmentA.6.2.4 AI system verification and validationMAP 5 Impacts on individuals, groups and societyMEASURE 3 Tracking identified AI risks over timeMANAGE 1 Prioritising and responding to AI risks
Document training, validation and testing data governanceArticle 10A.7.4 Quality of data for AI systemsA.7.5 Data provenanceA.7.6 Data preparationMAP 2 Categorisation of the AI systemMEASURE 2 Evaluation for trustworthy characteristics
Maintain technical documentation and instructions for useArticles 11 and 13; Annex IVA.6.2.7 AI system technical documentationA.8.2 System documentation and information for usersA.6.2.3 Documentation of AI system design and developmentMAP 2 Categorisation of the AI systemMAP 3 Capabilities, intended use, benefits and costs
Design logging and effective human oversightArticles 12 and 14A.6.2.8 AI system recording of event logsA.6.1.3 Processes for responsible AI system design and developmentMAP 3 Capabilities, intended use, benefits and costsMANAGE 2 Maximising benefits and minimising negative impactsMEASURE 3 Tracking identified AI risks over time
Evidence accuracy, robustness and cybersecurityArticle 15A.6.2.4 AI system verification and validationA.6.2.2 AI system requirements and specificationMEASURE 1 Appropriate methods and metricsMEASURE 2 Evaluation for trustworthy characteristics
Prepare quality management, conformity and registrationArticles 16–18, 43 and 47–494.4 AI management system7.5 Documented information9.2 Internal auditGOVERN 1 Policies, processes and procedures for AI riskGOVERN 2 Accountability structures, roles and training
Operate post-market monitoring and serious-incident responseArticles 20, 72 and 73A.6.2.6 AI system operation and monitoring10.2 Nonconformity and corrective actionA.8.4 Communication of incidentsMANAGE 4 Risk treatment, response, recovery and communicationMEASURE 3 Tracking identified AI risks over time
Use the high-risk system according to instructions with competent oversightArticle 26(1)–(3)A.9.4 Intended use of the AI systemA.9.2 Processes for responsible use of AI systems7.2 CompetenceGOVERN 2 Accountability structures, roles and trainingMAP 3 Capabilities, intended use, benefits and costs
Control input data, monitor operation and retain logsArticle 26(4)–(6)A.6.2.6 AI system operation and monitoringA.6.2.8 AI system recording of event logsA.7.4 Quality of data for AI systemsMANAGE 4 Risk treatment, response, recovery and communicationMEASURE 3 Tracking identified AI risks over timeMANAGE 2 Maximising benefits and minimising negative impacts
Tell workers and their representatives before the system is used at workArticle 26(7)A.8.5 Information for interested parties7.4 CommunicationGOVERN 5 Engagement with relevant AI actorsMAP 5 Impacts on individuals, groups and society
Tell people when the system helps make decisions about themArticle 26(11)A.8.5 Information for interested parties7.4 CommunicationMEASURE 2 Evaluation for trustworthy characteristicsGOVERN 5 Engagement with relevant AI actors
Be ready to explain decisions to the people affectedArticle 86A.8.5 Information for interested partiesA.8.3 External reportingMEASURE 2 Evaluation for trustworthy characteristicsMANAGE 4 Risk treatment, response, recovery and communication
Complete the impact assessments the deployment requiresArticles 26(9) and 276.1.4 AI system impact assessmentA.5.4 Assessing AI system impact on individuals or groups of individualsA.5.3 Documentation of AI system impact assessmentsMAP 5 Impacts on individuals, groups and societyMAP 1 Context is established and understoodMEASURE 2 Evaluation for trustworthy characteristics
Public authorities: register your use of the systemArticles 26(8) and 49(3)A.8.5 Information for interested partiesA.10.3 SuppliersGOVERN 1 Policies, processes and procedures for AI riskGOVERN 6 Third-party software, data and supply chain risks
Put written agreements in place with suppliers of AI componentsArticle 25(4)A.10.3 SuppliersA.10.2 Allocating responsibilitiesGOVERN 6 Third-party software, data and supply chain risksMANAGE 3 Third-party AI risks and benefitsMAP 4 Risks and benefits of all system components
Appoint an authorised representative in the EUArticle 22A.10.2 Allocating responsibilities7.5 Documented informationGOVERN 2 Accountability structures, roles and trainingGOVERN 1 Policies, processes and procedures for AI risk
Carry out the authorised representative's mandateArticle 22(3)–(4)7.5 Documented informationA.10.2 Allocating responsibilitiesA.8.5 Information for interested partiesGOVERN 1 Policies, processes and procedures for AI riskGOVERN 6 Third-party software, data and supply chain risks
Complete importer checks before placing the system on the marketArticle 23A.10.3 Suppliers8.1 Operational planning and controlGOVERN 6 Third-party software, data and supply chain risksMANAGE 3 Third-party AI risks and benefits
Complete distributor checks before making the system availableArticle 24A.10.3 Suppliers8.1 Operational planning and controlGOVERN 6 Third-party software, data and supply chain risksMANAGE 3 Third-party AI risks and benefits
Obtain and keep your model provider's informationArticle 53(1)(b); Annex XIIA.10.3 SuppliersA.4.2 Resource documentationMAP 4 Risks and benefits of all system componentsGOVERN 6 Third-party software, data and supply chain risksMANAGE 3 Third-party AI risks and benefits
Maintain GPAI technical documentation and downstream informationArticle 53(1)(a)–(b); Annexes XI and XIIA.6.2.7 AI system technical documentationA.8.2 System documentation and information for usersMAP 2 Categorisation of the AI systemMAP 3 Capabilities, intended use, benefits and costsMEASURE 2 Evaluation for trustworthy characteristics
Implement copyright policy and publish training-content summaryArticle 53(1)(c)–(d)A.7.3 Acquisition of dataA.7.5 Data provenanceA.8.5 Information for interested partiesMAP 4 Risks and benefits of all system componentsGOVERN 6 Third-party software, data and supply chain risksGOVERN 1 Policies, processes and procedures for AI risk

Working towards ISO/IEC 42001?

The crosswalk shows where your EU AI Act records already help. A certification audit still needs the full management system the standard describes, checked by an accredited body. See ISO 42001 and the EU AI Act.