One register entry, every register.

Fill one entry for an AI system or an agent. It is checked against the published schema, kept in a register in this browser, exported as JSON, CSV or a Markdown public summary, and mapped field by field to the UK ATRS, a Canada AIA, the EU database, a model card and an ISO/IEC 42001 SoA.

Indicative, not legal advice and not a conformity claim. Nothing you enter leaves your browser.

JavaScript is off or has not loaded, so the form, the checks, the register and the exports are not available. The questions, the criteria and the guide below still work as a worksheet you can read and fill in by hand.

Built from 04. The stack (five layers) of the Body of Knowledge. Runs entirely in this page: no account and no upload.

How to use it

  1. Choose what you are registering: an AI system or an agent. Each kind has its own published schema, and the entry is checked against it. The entry is the anchor every other record of the system is filed against, the inventory of Layer 02 [9].
  2. Fill the core first: id, version, owner, scope and expiry are required. The deploy pipeline is the natural writer of those fields; intake adds the classification.
  3. Fill the public record once if the system appears in any public register: the crosswalk below shows where each of its fields goes.
  4. Choose "Check the entry". Problems are listed at the top, each linked to its field. Then add the entry to the register, download it, or start the next one.

By hand, without JavaScript

Use the human templates, which name every field with its description: ai system entry template (schema: ai-system-register-entry.v1.json ) and agent entry template (schema: agent-register-entry.v1.json ) . The field lists below follow the form's sections.

AI system entry: 42 fields in 5 sections

Identity and ownership

  • Registry id (required): id
  • Version (required): version
  • Name : name
  • Accountable owner (team or role) (required): owner
  • Entry expires on (required): expiry
  • Last reviewed on : last_reviewed
  • Use-case record id : use_case_record

Purpose and scope

  • Intended purpose : purpose
  • Declared scope (required): scope
  • Kind of AI artefact : system_type
  • Lifecycle stage : lifecycle_stage
  • People affected : affected_persons
  • Where it is used : jurisdictions
  • Processes personal data : personal_data
  • Human oversight mode : human_oversight

Classification and registration

  • Role under the EU AI Act : role
  • EU AI Act category : risk_classification.eu_ai_act_category
  • Annex III point relied on : risk_classification.annex_iii_point
  • Internal risk tier : risk_classification.internal_tier
  • Why this classification : risk_classification.rationale
  • Signed classification decision (link) : risk_classification.classification_record
  • Decided on : risk_classification.decided_at
  • Registration in the EU database applies : eu_database_registration.required
  • EU database registration id : eu_database_registration.registration_id
  • Registered on : eu_database_registration.registered_at

Components and evidence

  • Components : components
  • AIBOM (link) : aibom
  • Evidence links : evidence

Public record

  • Responsible organisation : public_record.organisation
  • Contact email (a role mailbox) : public_record.contact_email
  • Address and other contact details : public_record.address
  • Public web page : public_record.website_url
  • One-sentence description : public_record.one_sentence_description
  • Plain-language description : public_record.description
  • Information used (data, inputs) and operating logic : public_record.data_and_logic
  • Trade name and unambiguous reference : public_record.trade_name
  • Market status : public_record.market_status
  • Region whose public it affects : public_record.region
  • Authorised representative : public_record.authorised_representative
  • Notified-body certificate : public_record.certificate
  • EU declaration of conformity (link) : public_record.declaration_of_conformity
  • Electronic instructions for use (link) : public_record.instructions_for_use
Agent entry: 46 fields in 6 sections

Identity and ownership

  • Agent id (required): id
  • Version (required): version
  • Name : name
  • Accountable owner (team) (required): owner
  • Entry expires on (required): expiry
  • Status : status
  • First registered at : registered_at

Purpose, scope and autonomy

  • What the agent is for : purpose
  • Declared scope (required): scope
  • Parent AI system (registry id) : parent_system
  • Base models : base_models
  • Highest autonomy allowed : autonomy_level
  • Action log location and retention : logging

Workload identity

  • Identity type : workload_identity.type
  • Identifier : workload_identity.identifier
  • Issuer : workload_identity.issuer
  • Credential lifetime (seconds) : workload_identity.credential_ttl_seconds

Tools, data and delegation

  • Tools : tools
  • Data access : data_access
  • May create sub-agents : delegation.can_spawn_agents
  • May delegate to (agent ids) : delegation.can_delegate_to

Oversight, limits and kill switch

  • Oversight mode : human_oversight.mode
  • Actions that always need approval : human_oversight.approval_required_for
  • Overseer role : human_oversight.overseer_role
  • Spend limit currency (ISO 4217) : spend_limit.currency
  • Spend limit amount : spend_limit.amount
  • Spend limit period : spend_limit.period
  • Kill switch: how the agent is stopped : kill_switch.mechanism
  • Kill switch: who can pull it : kill_switch.owner
  • Kill switch: last tested on : kill_switch.last_drill
  • Runtime policy cards (ids) : runtime_policies
  • Eval suites that gate releases (ids) : eval_suites

Public record

  • Responsible organisation : public_record.organisation
  • Contact email (a role mailbox) : public_record.contact_email
  • Address and other contact details : public_record.address
  • Public web page : public_record.website_url
  • One-sentence description : public_record.one_sentence_description
  • Plain-language description : public_record.description
  • Information used (data, inputs) and operating logic : public_record.data_and_logic
  • Trade name and unambiguous reference : public_record.trade_name
  • Market status : public_record.market_status
  • Region whose public it affects : public_record.region
  • Authorised representative : public_record.authorised_representative
  • Notified-body certificate : public_record.certificate
  • EU declaration of conformity (link) : public_record.declaration_of_conformity
  • Electronic instructions for use (link) : public_record.instructions_for_use

One record, five regimes

Each regime asks for much the same facts under its own names. Write them once in the register, then copy them across with this crosswalk. Field names are copied from each source as of 2026-09-24; an empty cell means the regime has no field for it. The mapping is illustrative, not a claim of conformity.

  • UK ATRS, template v4.0 [3]: Tier 1 (Section 1, Summary information, for the general public) has four fields, 1.1 Name, 1.2 Description, 1.3 Website URL and 1.4 Contact email; Section 0 holds the record's metadata (0.1 Title to 0.5 One sentence description); Tier 2 (sections 2.1 to 2.5) is for specialist readers.
  • Canada AIA [4]: the Project Details questions and the section names of the questionnaire. Its answers produce an impact level (I to IV) under the Directive on Automated Decision-Making [5], which is the regime's own classification.
  • EU database, Annex VIII of the EU AI Act [1]: Section A for providers of high-risk systems (Art. 49(1)), Section B for providers relying on the Art. 6(3) filter (Art. 49(2)) and Section C for deployers that register under Art. 49(3). The Digital Omnibus deleted Section B points 7 (summary of grounds) and 9 (Member States) [2].
  • Model or system card: the Hugging Face model card template headings [6] and the CycloneDX 1.7 model card fields [7]; the model card builder writes both.
  • ISO/IEC 42001 Statement of Applicability [8]: the SoA lists the Annex A controls with the justification for including or excluding each and whether it is implemented (verify the clause wording). A register entry does not make an SoA; it supplies the evidence and the justification for the rows named in the last column.
Register field to regime field
Register field UK ATRS v4.0Canada AIAEU database (Annex VIII)Model or system cardISO/IEC 42001 SoA (verify)
Name name 0.1 Title; 1.1 NameProject TitleA.4, B.4 trade name and unambiguous referenceHF "Model Card for {model_id}"; CycloneDX component nameNo direct field
Registry id id No direct fieldProject ID from IT PlanA.4, B.4 reference allowing identification and traceabilityCycloneDX component bom-refNo direct field
Version version 2.4.2.2 Model versionNo direct fieldA.4 (part of the unambiguous reference)CycloneDX component versionNo direct field
One-sentence description public_record.one_sentence_description 0.5 One sentence descriptionNo direct fieldNo direct fieldHF model summary (the line under the title)No direct field
Plain-language description public_record.description 1.2 Description; 2.2.1 Detailed descriptionProvide a project descriptionA.5 description of the intended purpose and of the components and functionsHF "Model Description"; CycloneDX component descriptionNo direct field
Intended purpose purpose 1.2 Description (what the tool is, why it is used)Reasons for Automation (section)A.5, B.5 intended purposeHF "Direct Use"; CycloneDX modelCard.considerations.useCasesA.9 row (intended use): justification
Responsible organisation public_record.organisation 0.2 Organisation name; 2.1.1 Organisation or departmentDepartmentA.1, B.1 provider; C.1 deployer (name, address, contact details)HF "Developed by"; CycloneDX component supplierNo direct field
Address and contact details public_record.address No direct fieldNo direct fieldA.1, B.1, C.1 (address and contact details); A.2, B.2, C.2 for whoever submits on the provider's or deployer's behalfNo direct fieldNo direct field
Contact email public_record.contact_email 1.4 Contact emailNo direct fieldA.1, B.1, C.1 (contact details)HF "Model Card Contact"No direct field
Public web page public_record.website_url 1.3 Website URLNo direct fieldA.13 URL for additional information (optional)HF "Model Sources"; CycloneDX externalReferences (website)No direct field
Accountable owner owner 2.1.2 Team; 2.1.3 Senior responsible ownerName of ADM responsible for the program; BranchNo direct fieldHF "Model Card Authors"A.3 row (roles and responsibilities): owner
Lifecycle stage lifecycle_stage 0.3 PhaseProject Phase (Design or Implementation)No direct fieldNo direct fieldA.6 row (life cycle): implementation status
Market status public_record.market_status No direct fieldNo direct fieldA.7, B.8 status of the AI systemNo direct fieldNo direct field
Region public_record.region 0.4 RegionNo direct fieldNo direct fieldNo direct fieldNo direct field
Where it is used jurisdictions No direct fieldNo direct fieldA.10 Member States (the B.9 list was deleted by Reg. (EU) 2026/1744)HF "Language(s)" only for languages, not placesNo direct field
EU AI Act classification risk_classification.eu_ai_act_category + risk_classification.annex_iii_point No direct fieldImpact level I to IV, from the Risk Profile and Impact Assessment answersSection A (high-risk) or Section B (Art. 6(3) filter); B.6 conditions relied onNo direct fieldJustification for inclusion or exclusion of controls
Why this classification risk_classification.rationale No direct fieldNo direct fieldB.6 (the B.7 summary of grounds was deleted by Reg. (EU) 2026/1744)No direct fieldJustification for inclusion or exclusion of controls
Information used and operating logic public_record.data_and_logic 2.4.1.2 System-level input; 2.4.1.3 System-level output; 2.4.3.1 Development data descriptionAbout the Data; About the Algorithm (sections)A.6 information used (data, inputs) and operating logicHF "Training Data"; CycloneDX modelCard.modelParameters (datasets, inputs, outputs)A.7 row (data for AI systems): evidence
Components and AIBOM components + aibom 2.4.1.5 Models; 2.1.4 Third party involvementAbout The System: "Who developed the system?"No direct fieldCycloneDX components (the ML-BOM itself)A.4 row (resources) and A.10 row (suppliers): evidence
Human oversight mode human_oversight 2.3.2 Human reviewDe-risking and Mitigation Measures (section)A.12 via the instructions for useSystem card: human oversight measures (Art. 13(3)(d))A.9 row (use of AI systems): evidence
People affected affected_persons 2.5.2 Risks and mitigationsImpact Assessment (section)C.4 summary of the FRIA findings (deployers)CycloneDX modelCard.considerations.fairnessAssessments (groupAtRisk)A.5 row (impact assessment): evidence
Processes personal data personal_data 2.4.3.4 Sensitive attributes; 2.4.4.2 Sensitive attributesAbout the Data (section)C.5 summary of the DPIA (deployers, where applicable)No direct fieldA.7 row (data for AI systems): justification
Trade name public_record.trade_name No direct fieldNo direct fieldA.4, B.4 trade nameNo direct fieldNo direct field
Authorised representative public_record.authorised_representative No direct fieldNo direct fieldA.3, B.3 authorised representativeNo direct fieldNo direct field
Notified-body certificate public_record.certificate No direct fieldNo direct fieldA.8 certificate type, number, expiry and notified body; A.9 scanned copyNo direct fieldNo direct field
EU declaration of conformity public_record.declaration_of_conformity No direct fieldNo direct fieldA.11 copy of the EU declaration of conformityNo direct fieldNo direct field
Instructions for use public_record.instructions_for_use No direct fieldNo direct fieldA.12 electronic instructions for use (not for Annex III points 1, 6 and 7)System card: instructions for use (Art. 13)A.8 row (information for interested parties): evidence
EU database registration id eu_database_registration.registration_id No direct fieldNo direct fieldThe registration itself; C.3 URL of the provider's entry (deployers)No direct fieldNo direct field
Evidence links evidence 2.5.1 Impact assessmentsNo direct fieldC.4 FRIA summary and C.5 DPIA summary (deployers)HF "More Information"; CycloneDX externalReferencesImplementation status and evidence of the applicable controls

Bulk import and export

  • JSON. One record, a list of records, or an object with an entries list. A record's $schema says which kind it is; without one, agent-only fields (workload identity, tools, kill switch) mark an agent.
  • CSV (RFC 4180). The header row holds the field names, nested ones with dots (risk_classification.internal_tier, public_record.contact_email); an optional kind column says system or agent. A list of text goes in one cell, separated by ; ; a list of records (components, tools, data access) goes in one cell as JSON. Booleans read true/false or yes/no.
  • The register CSV export uses the same shape, so an export imports back unchanged. Exported text that a spreadsheet would run as a formula gets a leading apostrophe, which the import removes.
  • Every imported entry is checked: the import report lists each entry that does not validate, with its first problems, and the register table shows the count for every entry.

What this is not

It is not a registration in the EU database, an ATRS submission or a completed AIA: each regime's own form and process decide. The crosswalk says where a field goes; it does not say that a regime applies to your system. Drafts stay in this browser and are not in the link you copy, because an entry can hold internal detail. The mappings are illustrative, not a claim of conformity.

Sources

  1. [1] Regulation (EU) 2024/1689 (Artificial Intelligence Act), Art. 49 (registration) and Annex VIII, Sections A (providers of high-risk systems, points 1 to 13), B (systems relying on Art. 6(3), points 1 to 9) and C (deployers, points 1 to 5), consolidated text of 2026-07-27. EUR-Lex refused automated access on 2026-09-24; the wording was read on the Commission's AI Act Service Desk, which reproduces the Official Journal text. Publications Office of the EU (EUR-Lex). 2026-07-27. https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng#anx_VIII (verified: primary)
  2. [2] Regulation (EU) 2026/1744 (Digital Omnibus on AI), deleting Annex VIII, Section B, points 7 (summary of grounds) and 9 (Member States). Publications Office of the EU (EUR-Lex). 2026-07-24. https://eur-lex.europa.eu/eli/reg/2026/1744/oj/eng (verified: primary)
  3. [3] Algorithmic Transparency Recording Standard (ATRS) template, version 4.0: Section 0 metadata (0.1 Title to 0.5 One sentence description), Tier 1 Section 1 Summary (1.1 Name, 1.2 Description, 1.3 Website URL, 1.4 Contact email) and the Tier 2 sections 2.1 to 2.5. Department for Science, Innovation and Technology, GOV.UK. 2025-05-08. https://www.gov.uk/government/publications/algorithmic-transparency-template (verified: primary)
  4. [4] Algorithmic Impact Assessment tool, questionnaire source (survey-enfr.json, version 1.0.1): Project Details fields (Project Title, Project ID from IT Plan, Department, Branch, Name of ADM responsible for the program, Project Phase, project description) and the section names. Government of Canada, canada-ca/aia-eia-js on GitHub. 2025-10-03. https://github.com/canada-ca/aia-eia-js (verified: primary)
  5. [5] Directive on Automated Decision-Making (an Algorithmic Impact Assessment completed and published before production; impact levels I to IV). Treasury Board of Canada Secretariat. 2025-06-24. https://www.tbs-sct.canada.ca/pol/doc-eng.aspx?id=32592 (verified: primary)
  6. [6] Model card template (section headings: Model Details, Uses, Bias, Risks, and Limitations, Training Details, Evaluation, Environmental Impact, Model Card Contact). Hugging Face, huggingface_hub. 2026. https://github.com/huggingface/huggingface_hub/blob/main/src/huggingface_hub/templates/modelcard_template.md (verified: primary)
  7. [7] CycloneDX specification 1.7, JSON reference (component of type machine-learning-model; modelCard with modelParameters, quantitativeAnalysis and considerations). OWASP CycloneDX, Ecma TC54. 2025-10-21. https://cyclonedx.org/docs/1.7/json/ (verified: primary)
  8. [8] ISO/IEC 42001:2023, AI management systems (Annex A control areas A.2 to A.10; the Statement of Applicability of clause 6.1.3). ISO/IEC. 2023-12. The SoA column wording was not opened: verify it against the standard. https://www.iso.org/standard/81230.html (verified: primary)
  9. [9] 04. The stack, Layer 02: Inventory & Transparency, and the Agent Registry pattern: the registry entry this builder writes. AI Governance Engineering Body of Knowledge. 2026-09-24. https://aigovernanceengineer.com/bok/the-stack#layer-02-inventory--transparency (verified: primary)