Annex IV documentation assembly and consistency protocol (PR-11-DOC)
PR-11-DOC checks the technical documentation of a high-risk AI system against every sub-point of Annex IV, assembles missing sections from test results, and compares each statement in the instructions for use under Art. 13(3) with the measured performance recorded under Art. 15. For a general-purpose AI model it runs the same checks against Annex XI and Annex XII. It produces a gap and consistency report, assembled section drafts with a source map, and an obligation to test to evidence traceability matrix.
| Field | Value |
|---|---|
| Protocol | PR-11-DOC |
| Version | v1.0 |
| Articles | |
| Derived from |
|
PR-11-DOC v1.0 is built from Annex IV of Regulation (EU) 2024/1689, Art. 13(3), the simplified technical documentation form under Art. 11(1) third subparagraph where the Commission has issued it, Annex XI and Annex XII of Regulation (EU) 2024/1689, and the Transparency chapter of the GPAI Code of PracticePublished with its Model Documentation Form. The format in which each declared metric is stated follows prEN 18229-2Draft, and the document control routine that keeps the assembled set up to date follows prEN 18286Published.
Scope
PR-11-DOC applies to a high-risk AI system under Annex III or Annex I that needs technical documentation under Art. 11 and instructions for use under Art. 13, and to a general-purpose AI model whose provider needs technical documentation under Art. 53(1) with the Annex XI content and downstream information with the Annex XII content. Annex III systems fall under these duties from the Annex III application date Deferred; general-purpose AI model duties apply from the GPAI application date In force.
The protocol covers three groups of obligation elements. The first group is completeness: every sub-point of Annex IV sections 1 to 9, every point of Annex XI and Annex XII, and every item of Art. 13(3)(a) to (f) is present, partial, absent or marked not applicable, with a location. The second group is assembly: sections that a test result can supply, such as Annex IV section 2(g), section 3, section 4 and section 7, are drafted from the test reports and records with a source reference on each statement. The third group is consistency: every declared value, version, component, dataset description, oversight measure and security measure in the documentation matches the record it rests on, and every statement in the instructions for use matches the technical documentation and the measured performance.
The protocol does not run accuracy, robustness, cybersecurity or bias tests; those results come from PR-15-ROB v1.0 and PR-10-BIAS v1.0, or from the provider's own dated and signed reports. It does not verify the oversight functions themselves; that is PR-14-OVR v1.0. It does not determine whether a modifier is the provider of a general-purpose AI model; that is PR-CLS v1.0. It does not judge whether the documentation is clear and comprehensive in the sense of Art. 11(1); that judgement rests with the national competent authority or the notified body. It does not assess the content of the copyright policy under Art. 53(1)(c) beyond its presence. The service boundaries are stated on what we cannot do.
Articles covered
| Article | Testable element | Steps |
|---|---|---|
| Art. 11 | The accuracy, robustness and other metric values declared in the documentation match the values in the dated and signed test reports it contains, for the same system version and the same test conditions (Annex IV section 2(g)) | 16, 20 |
| Art. 11 | The degrees of accuracy stated for specific persons or groups of persons and the overall expected level of accuracy match the disaggregated and aggregate results in the test reports (Annex IV section 3) | 21 |
| Art. 11 | The system version and the software or firmware versions stated in the general description match the deployed artefact and the versions named in the test reports (Annex IV section 1(a) and 1(c)) | 28 |
| Art. 11 | The architecture description matches the system as built: the listed components, their interfaces and the integrations with external hardware, software and other AI systems are present and no deployed component is missing from the description (Annex IV section 2(c) and 1(b)) | 29 |
| Art. 11 | The intended purpose and the conditions of use in the general description match the instructions for use and the promotional or sales materials (Annex IV section 1(a) and 1(h)) | 27 |
| Art. 11 | The datasheets match the data sets actually used: provenance, scope, main characteristics, labelling procedure and cleaning methodology as recorded in the data pipeline (Annex IV section 2(d)) | 30 |
| Art. 11 | The human oversight measures and the output interpretation measures described in the documentation are present in the system and in the instructions for use (Annex IV section 2(e) and section 3) | 31 |
| Art. 11 | The cybersecurity measures described in the documentation are present in the deployed configuration (Annex IV section 2(h)) | 32 |
| Art. 11 | Every test log and test report in the documentation carries a date and the signature of a responsible person, and the set covers each pre-determined change described in section 2(f) (Annex IV section 2(g)) | 33 |
| Art. 11 | The lifecycle change record matches the version history in the general description and the versions covered by test reports (Annex IV section 6 with section 1(a) and 1(c)) | 34 |
| Art. 11 | Each entry in the standards and specifications list maps to the requirement it addresses, and each Chapter III Section 2 requirement without an applied harmonised standard maps to a described solution (Annex IV section 7) | 18 |
| Art. 11 | The copy of the EU declaration of conformity names the same system, version and provider as the general description (Annex IV section 8 with section 1(a)) | 35 |
| Art. 13 | The accuracy level and metrics declared in the instructions for use under Art. 13(3)(b)(ii) match the result of a test on a held-out set under the declared conditions of use. | 22 |
| Art. 13 | The robustness and cybersecurity levels declared in the instructions for use under Art. 13(3)(b)(ii) match the result of perturbation and adversarial tests under the declared conditions of use. | 23 |
| Art. 13 | The known and foreseeable circumstances declared under Art. 13(3)(b)(ii) and (iii) reproduce the declared change in performance or the declared risk when the system runs under those circumstances. | 24 |
| Art. 13 | The performance declared for specific persons or groups under Art. 13(3)(b)(v) matches the disaggregated results recorded in the technical documentation. | 25 |
| Art. 13 | Each statement in the instructions for use on intended purpose, input data specifications, pre-determined changes, human oversight measures and log mechanisms (Art. 13(3)(b)(i), (b)(vi), (c), (d) and (f)) matches the corresponding section of the Annex IV technical documentation. | 12, 26 |
| Art. 53 | Provider status after modification: whether a fine-tuning, continued training or other modification makes the modifier the provider of a new general-purpose AI model, determined against the criteria in the Commission guidelines on GPAI scope, including the compute used for the modification relative to the training compute of the original model | 4 |
| Art. 53 | Art. 53(2) exception status: whether the licence allows access, usage, modification and distribution, and whether the parameters including the weights, the architecture information and the usage information are publicly accessible at the stated locations | 4 |
| Art. 53 | Annex XI completeness: every element of Section 1 point 1(a) to (f) and point 2(a) to (e), and of Section 2 for a model with systemic risk, is present in the technical documentation and traceable to a record | 8, 11, 38 |
| Art. 53 | Annex XII completeness: every element of point 1(a) to (h) and point 2(a) to (c) is present in the information made available to downstream AI system providers | 8, 11 |
| Art. 53 | Cross-document consistency: the architecture, number of parameters, modality, format, release date, distribution method and licence match between the Annex XI documentation, the Annex XII information and the public model card | 36 |
| Art. 53 | Declared input and output limits: the modality, format and maximum input and output size stated under Annex XII point 2(b) match the behaviour observed on a sample of inputs at and beyond the stated limits | 37 |
| Art. 53 | Evaluation result traceability: each evaluation result cited under Art. 53(1)(a) carries the evaluation protocol name, the dataset version and the model version, and each identifier resolves to a stored record | 38 |
| Art. 53 | Training-content summary structure: the published summary follows the sections of the AI Office template and each section carries content rather than a placeholder | 39 |
The first two Art. 53 elements are determined under PR-CLS v1.0; step 4 takes the two determination records as inputs and fixes which Art. 53(1) duties the completeness checks cover. The obligation pages for Art. 11, Art. 13 and Art. 53 give the documentation and process elements that sit outside this protocol.
Inputs required
| Input | Format | Provided by |
|---|---|---|
| Technical documentation as it stands, including any draft Annex IV set or Commission simplified form | Document set with version identifiers, any file format that preserves text | Provider |
| Instructions for use as supplied to deployers | Document with version identifier, in the digital format of delivery | Provider |
| Promotional and sales materials that state the intended purpose or conditions of use | Documents, web page exports or presentations, with capture identifiers | Provider |
| Test reports and test logs for accuracy, robustness, cybersecurity and bias | PR-15-ROB and PR-10-BIAS outputs, or the provider's own reports, each dated and signed | Provider or AIAssuranceLabs from earlier protocols |
| Human oversight verification record | PR-14-OVR output, or the provider's own record of oversight measures | Provider or AIAssuranceLabs from an earlier protocol |
| Deployed artefact identity | Build identifier, container digest or package version of the system under review | Provider |
| Component inventory and deployed configuration | Software bill of materials, architecture diagram, configuration export, external integration list | Provider |
| Data pipeline records for training, validation and testing data sets | Dataset manifests, provenance records, labelling procedure, cleaning scripts or logs | Provider |
| Risk management file, lifecycle change record, post-market monitoring plan | Documents with version identifiers | Provider |
| Copy of the EU declaration of conformity, where drawn up | Document | Provider |
| For a general-purpose AI model: Annex XI documentation, Annex XII information pack, public model card, published training-content summary | Documents with version identifiers; URL of the public card and summary | Provider |
| For a general-purpose AI model: provider determination record and open-source exception record | PR-CLS outputs | AIAssuranceLabs from an earlier protocol |
| For a general-purpose AI model: query access to the model at its declared input and output limits | API endpoint or local runtime with the declared maximum sizes | Provider |
| Frozen input set | Inventory with one identifier and one content hash per document version | AIAssuranceLabs, generated in step 6 |
Steps
Preparation
Record the system name, provider, version under review and the deployed build identifier.
Record System identity record: name, provider, version under review, deployed build identifier
[AIAssuranceLabs procedure]
Determine the documentation route: full Annex IV, the Art. 11(1) simplified form, or Annex XI and XII.
Record Route determination record; where the Commission has issued the simplified form, its version
[Regulation (EU) 2024/1689, Art. 11(1) third subparagraph and Art. 53(1)]
For an Annex I Section A product, list the product legislation content that joins the single set.
Record Single documentation set scope record naming the product legislation content included
[Regulation (EU) 2024/1689, Art. 11(2)]
For a general-purpose AI model, take the provider determination and open-source exception records from
PR-CLS.Record Art. 53 scope record stating which of Art. 53(1)(a) and (b) apply, with the two PR-CLS record identifiers
[Regulation (EU) 2024/1689, Art. 53(1) and Art. 53(2)]
Collect the technical documentation, instructions for use, promotional materials, test reports, test logs and data pipeline records.
Record Input inventory listing each document with its version identifier
[AIAssuranceLabs procedure]
Assign an identifier and a content hash to each document version in the inventory.
Record Frozen input set: one identifier and one content hash per document version
[AIAssuranceLabs procedure]
Build the gap checklist with one row for each sub-point of Annex IV sections 1 to 9.
Record Annex IV checklist template with one row per sub-point
[Regulation (EU) 2024/1689, Annex IV sections 1 to 9]
For a general-purpose AI model, build one row for each Annex XI and XII point, mapped to its Model Documentation Form field.
Record Annex XI and XII checklist template with the Model Documentation Form field named on each row
[Regulation (EU) 2024/1689, Annex XI and Annex XII; GPAI Code of Practice, Transparency chapter, Measure 1.1 and Model Documentation Form]
Build the instructions-for-use checklist with one row for each item in Art. 13(3)(a) to (f).
Record Instructions-for-use checklist template with one row per item in Art. 13(3)(a) to (f), including (b)(i) to (b)(vii)
[Regulation (EU) 2024/1689, Art. 13(3)]
Execution
Mark each Annex IV row present, partial, absent or not applicable, with the location of the content.
Record Annex IV gap table: status and location per row
[Regulation (EU) 2024/1689, Annex IV, introductory sentence]
Mark each Annex XI and XII row present, partial or absent, with the location of the content.
Record Annex XI and XII gap table: status, location and record reference per row; the provider's reason for each row marked not applicable
[Regulation (EU) 2024/1689, Art. 53(1)(a) and Art. 53(1)(b)]
Mark each Art. 13(3) row present, partial or absent in the instructions for use.
Record Instructions-for-use completeness checklist against Art. 13(3)(a) to (f)
[Regulation (EU) 2024/1689, Art. 13(3)]
For each absent or partial row, name the test report or record that supplies the content.
Record Source map: gap row to supplying record, or 'no record' where none exists
[AIAssuranceLabs procedure]
Draft section 2(g) from the test reports: procedures, test data, metrics, logs and signatures.
Record Section 2(g) draft with a source reference on each statement
[Regulation (EU) 2024/1689, Annex IV section 2(g)]
Draft section 3 from the test reports: capabilities, limitations and degrees of accuracy per group.
Record Section 3 draft with a source reference on each statement
[Regulation (EU) 2024/1689, Annex IV section 3]
State each declared metric with its name, value, test set, conditions and system version.
Record Metric declaration entries in the fixed format, with any rounding rule; a value without a test set or conditions is marked incomplete
[prEN 18229-2, draft, declaration structure for accuracy and robustness levels]
Draft section 4 from the provider's written justification of each metric against the context of use.
Record Section 4 draft; each metric without a written justification is marked
[Regulation (EU) 2024/1689, Annex IV section 4]
Draft section 7: each standard with its stage and requirement, and the described solution where no harmonised standard applies.
Record Standards and specifications list with requirement mapping; harmonised standards cited in the Official Journal listed separately, none as of the anchor date
[Regulation (EU) 2024/1689, Annex IV section 7]
Draft sections 5, 6 and 9 from the risk management file, change record and post-market monitoring plan.
Record Section 5, 6 and 9 drafts with source references
[Regulation (EU) 2024/1689, Annex IV sections 5, 6 and 9]
Compare each declared metric value in sections 2(g) and 3 with the test report for the same version and conditions.
Record Declared metric to test report consistency table: one row per declared value, compared against the unrounded report value
[Regulation (EU) 2024/1689, Annex IV section 2(g)]
Compare the degrees of accuracy per group and overall with the disaggregated and aggregate results.
Record Rows in the consistency table for each group and for the aggregate
[Regulation (EU) 2024/1689, Annex IV section 3]
Compare the accuracy level and metrics in the instructions for use with the held-out test result.
Record Instructions-for-use to measured performance row: declared accuracy against the PR-15-ROB held-out result
[Regulation (EU) 2024/1689, Art. 13(3)(b)(ii)]
Compare the robustness and cybersecurity levels in the instructions for use with the perturbation and adversarial results.
Record Instructions-for-use to measured performance rows: declared robustness and cybersecurity against the PR-15-ROB results
[Regulation (EU) 2024/1689, Art. 13(3)(b)(ii)]
For each declared circumstance, compare the declared performance change or risk with the test recorded under that circumstance.
Record Circumstance row: declared effect, tested effect, or 'no test' referred to PR-15-ROB
[Regulation (EU) 2024/1689, Art. 13(3)(b)(ii) and (iii)]
Compare the group performance in the instructions for use with the disaggregated results in the technical documentation.
Record Group performance row: declared value against the disaggregated result
[Regulation (EU) 2024/1689, Art. 13(3)(b)(v)]
Compare each statement on purpose, input data, pre-determined changes, oversight and logs with its Annex IV section.
Record Instructions-for-use to technical documentation consistency report: one row per statement with its Annex IV section
[Regulation (EU) 2024/1689, Art. 13(3)(b)(i), (b)(vi), (c), (d) and (f)]
Compare the intended purpose and conditions of use with the instructions for use and promotional materials.
Record Intended purpose consistency row: general description, instructions for use, each promotional item
[Regulation (EU) 2024/1689, Annex IV section 1(a) and 1(h)]
Compare the system, software and firmware versions with the deployed artefact and the test reports.
Record Version and architecture reconciliation record: version rows
[Regulation (EU) 2024/1689, Annex IV section 1(a) and 1(c)]
Compare the architecture description with the deployed component inventory, interfaces and external integrations.
Record Version and architecture reconciliation record: component rows, with undocumented components listed
[Regulation (EU) 2024/1689, Annex IV section 2(c) and 1(b)]
Compare each datasheet with the data pipeline records: provenance, scope, characteristics, labelling and cleaning.
Record Datasheet consistency rows: one per data set and attribute
[Regulation (EU) 2024/1689, Annex IV section 2(d)]
Confirm that each described oversight and output interpretation measure is present in the system and the instructions.
Record Oversight measure presence check, referencing the PR-14-OVR record where one exists
[Regulation (EU) 2024/1689, Annex IV section 2(e) and section 3]
Confirm that each described cybersecurity measure is present in the deployed configuration.
Record Cybersecurity measure presence check against the configuration export
[Regulation (EU) 2024/1689, Annex IV section 2(h)]
Check each test log and report for a date, a responsible signature and the pre-determined change it covers.
Record Test report register: date, signatory, version, pre-determined change covered
[Regulation (EU) 2024/1689, Annex IV section 2(g)]
Compare the lifecycle change record with the version history and the versions the test reports cover.
Record Lifecycle change consistency rows: change record, version history, test report coverage
[Regulation (EU) 2024/1689, Annex IV section 6 with section 1(a) and 1(c)]
Compare the system, version and provider in the EU declaration of conformity copy with the general description.
Record Declaration of conformity consistency row
[Regulation (EU) 2024/1689, Annex IV section 8 with section 1(a)]
Compare architecture, parameters, modality, format, release date, distribution and licence across Annex XI, Annex XII and the model card.
Record Cross-document consistency table: Annex XI documentation, Annex XII information, public model card
[Regulation (EU) 2024/1689, Annex XI Section 1 point 1 and Annex XII point 1]
Record the observed behaviour for sample inputs at and beyond the declared maximum input and output size.
Record Input and output limit observation log: declared limit, sample input size, observed behaviour
[Regulation (EU) 2024/1689, Annex XII point 2(b)]
For each cited evaluation result, check that the protocol name, dataset version and model version resolve to a stored record.
Record Evaluation result traceability rows: protocol name, dataset version, model version, resolved record identifier
[Regulation (EU) 2024/1689, Art. 53(1)(a) and Annex XI]
Mark each AI Office template section as present with content, present as placeholder, or absent in the summary.
Record Training-content summary structure check: one row per AI Office template section
[Regulation (EU) 2024/1689, Art. 53(1)(d)]
Analysis
Classify each finding as absent, partial, inconsistent or unverifiable, with its Annex or Article reference.
Record Findings register with one row per finding
[AIAssuranceLabs procedure]
Record, for each declared value, the section 4 location of its justification or the absence of one.
Record Justification location record: declared value, section 4 location, or 'no justification'
[Regulation (EU) 2024/1689, Annex IV section 4]
Record the document control version, owner and update trigger for each assembled section.
Record Document control record per assembled section: version, owner, update trigger
[prEN 18286, published, clauses on document and record control]
Build the traceability matrix: one row for each testable element, with protocol, step, evidence and result.
Record Obligation to test to evidence traceability matrix
[AIAssuranceLabs procedure]
Reporting
Write the gap and consistency report from the gap tables, consistency tables and findings register.
Record Annex IV gap and consistency report; for a general-purpose AI model, the Annex XI and XII completeness and consistency report
[AIAssuranceLabs procedure]
Deliver the assembled section drafts with the source map for the provider's review and signature.
Record Delivery record listing each section draft, its source map and the frozen input set identifier
[AIAssuranceLabs procedure]
The consistency table produced in step 20 holds one row per declared value. The row layout below is synthetic; the values belong to no system.
row_id | DMC-0007
doc_location | Annex IV section 2(g), table 3, row 2
metric_name | balanced_accuracy
declared_value | 0.91
declared_set | held-out-set-v3
declared_conditions | indoor lighting, single face, frontal pose
declared_version | 2.4.0
report_id | TR-0012 (PR-15-ROB output)
report_value | 0.91
report_set | held-out-set-v3
report_conditions | indoor lighting, single face, frontal pose
report_version | 2.4.0
report_dated_signed | yes
result | consistent
note | value, set, conditions and version identicalThe traceability matrix produced in step 43 holds one row per testable element of the covered obligation records. The row layout below is synthetic.
matrix_id | TM-0021
obligation | art-11
element | Annex IV section 2(g): declared metric values match dated and signed test reports
protocol | PR-11-DOC v1.0
steps | 16, 20
evidence_record | DMC-0007 (declared metric to test report consistency table)
source_record | TR-0012 (PR-15-ROB output)
input_hash | sha256:<hash of the frozen document version>
result | consistent | inconsistent | absent | unverifiable
finding_id | <findings register id, or none>
artifact | Annex IV gap and consistency report (PR-11-DOC output)Thresholds and criteria
Art. 11, Annex IV, Art. 13 and Art. 53 set no numeric threshold. The pass criterion for every step of PR-11-DOC is binary: an element is present or absent, a statement is traceable to a record or is not, and two statements are consistent or are not. No harmonised standard is cited in the Official Journal of the European Union as of the anchor date, so no presumption of conformity is available for any element this protocol checks.
A declared value is consistent with its test report when the metric name, the test set identifier, the test conditions and the system version are identical and the declared value equals the reported value or lies inside the reported interval. No numeric tolerance is applied, because no published source in the hierarchy gives one. Where the provider rounds a value, step 16 records the rounding rule as part of the declaration entry, and step 20 compares against the unrounded report value.
The hierarchy for the form of a declared value is, in order: a harmonised standard cited in the Official Journal (none as of the anchor date); a JTC 21 draft, which is prEN 18229-2Draft for the structure of accuracy and robustness declarations and prEN 18229-3Draft for the content of the instructions for use, with no public text at drafting stage; an ISO/IEC standard, of which none on the standards page covers Annex IV structure; a Commission code, which is the GPAI Code of PracticePublished Model Documentation Form for Annex XI and Annex XII fields; a sector framework; and the state of the art.
The declared values themselves come from the provider's testing under Art. 15. Their level is set by the provider against the context of use; the test-plan step of PR-15-ROB v1.0 records that justification when AIAssuranceLabs runs the test. PR-11-DOC does not set or judge the level. Step 41 records where each justification sits in the Annex IV section 4 statement and marks each declared value that has none. For a general-purpose AI model, the completeness target is every Annex XI and Annex XII point; the provider's reasons for marking a point not applicable are recorded in the gap table at step 11.
Outputs
| Artifact | Contents | Deliverable |
|---|---|---|
Annex IV gap and consistency report (PR-11-DOC output) | Annex IV gap table, findings register, consistency tables, justification location record, frozen input set identifier | Annex IV technical documentation gap analysis, Technical documentation and evidence-pack preparation |
Assembled Annex IV section drafts with source map (PR-11-DOC output) | Drafts of sections 2(g), 3, 4, 5, 6, 7 and 9 with a source reference on each statement; section 7 lists each standard with its stage | Assembled technical file sections drawn from test results, Technical documentation and evidence-pack preparation |
Declared metric to test report consistency table (PR-11-DOC output) | One row per declared value in the documentation and the instructions for use, with the report row it rests on | Consistency check between instructions for use (Article 13) and measured performance (Article 15), Technical documentation and evidence-pack preparation |
Instructions-for-use completeness checklist against Art. 13(3)(a) to (f) (PR-11-DOC output) | Status and location for each Art. 13(3) item | Consistency check between instructions for use (Article 13) and measured performance (Article 15), Technical documentation and evidence-pack preparation |
Instructions-for-use to technical documentation consistency report (PR-11-DOC output) | One row per statement on purpose, input data, pre-determined changes, oversight, logs and circumstances, with its Annex IV section and result | Consistency check between instructions for use (Article 13) and measured performance (Article 15), Technical documentation and evidence-pack preparation |
Version and architecture reconciliation record (PR-11-DOC output) | Version rows, component rows, undocumented components, lifecycle change rows, declaration of conformity row | Annex IV technical documentation gap analysis, Technical documentation and evidence-pack preparation |
Obligation to test to evidence traceability matrix (PR-11-DOC output) | One row per testable element with protocol, step, evidence record, source record, input hash and result | Obligation to test to evidence traceability matrix, Technical documentation and evidence-pack preparation |
Annex XI and XII completeness and consistency report (PR-11-DOC output) | Annex XI and XII gap table with Model Documentation Form fields, cross-document consistency table, input and output limit observation log, evaluation traceability rows, training-content summary structure check | Annex XI technical documentation contribution and Annex XII downstream information pack, GPAI evaluation support |
Each artifact is specified with a synthetic sample on the deliverables page.
Limitations
The protocol shows consistency, presence and traceability. It does not show that a declared value is correct; a value that matches its test report is only as reliable as that report, and a report the provider produced without an independent test is taken as given. It does not show that the documentation is clear and comprehensive to a national competent authority or a notified body, and it does not show that a notified body will accept the Commission simplified form in a given conformity assessment.
The protocol works on a frozen input set. A change to the system, the documentation or the instructions for use after step 6 is outside the result; step 42 records the update trigger for each section so that the provider can repeat the affected steps.
For a general-purpose AI model, the protocol checks presence and cross-document consistency of the Annex XI and Annex XII content. It does not judge whether the Annex XII information gives a downstream provider a good understanding of capabilities and limitations under Art. 53(1)(b), does not assess the copyright policy under Art. 53(1)(c), and does not verify the figures for compute or energy under Annex XI Section 1 points 2(d) and 2(e) beyond the presence of the records behind them. Where the Annex XII information of a model that generates content describes a marking technique under Art. 50(2), the protocol records the statement and its consistency with the technical documentation; the adequacy of the technique is tested under PR-50-MARK v1.0 and is an unsettled point.
Sources
- Regulation (EU) 2024/1689, consolidated text on EUR-Lex: https://eur-lex.europa.eu/eli/reg/2024/1689/oj, for Art. 11, Art. 13, Art. 53, Annex IV, Annex XI and Annex XII
GPAI Code of PracticePublished, the General-Purpose AI Code of Practice, Transparency chapter with the Model Documentation FormprEN 18229-2Draft, AI trustworthiness framework, part 2: accuracy and robustness, for the declaration structure of accuracy and robustness levelsprEN 18229-3Draft, AI trustworthiness framework, part 3: transparency and human oversight, for the content of the instructions for use once a text is publicprEN 18286Published, quality management system for EU AI Act regulatory purposes, for document and record controlCommission guidelines on GPAI scopePublished, guidelines on the scope of obligations for providers of general-purpose AI models, used byPR-CLSfor the two determination records that step 4 takes as inputs
Revision history
Each revision names the regulation entry whose verified-on date anchors the standards as they stood when the version was written.
| Version | Note | Anchor |
|---|---|---|
| v1.0 | Initial version, written from the standards in the derivation line as they stood on the verified-on date of the entry. | Verified on K&L Gates, EU Digital Omnibus on AI Enters Into Force (31 July 2026) |