AI Governance Engineer
GitHub

EU AI Act Art. 15(4): feedback loops in systems that continue to learn

Systems that continue to learn after placing on the market are built to eliminate or reduce the risk of biased outputs feeding future inputs (feedback loops), with mitigation measures

By Reviewed

From clause to evidenceThe chain from EU AI Act Art. 15(4) to its evidence record, in 6 steps; the text alternative lists them and the facts under the figure state each in full.Clause: EU AI Act (post-Omnibus), Art. 15(4)ClauseEU AI ActArt. 15Duty holder: ProviderDuty holderProviderApplies from: 2027-12-02 (Deferred)Applies from2027-12-02DeferredArtefact: Feedback-loop fairness monitor; retraining-data bias check; agent memory write gate with provenance and rollback to a known-good snapshotArtefactFeedback-loopfairness…Layers: Layer 03 Evals & Red Teaming as Evidence and Layer 04 Runtime Controls & ObservabilityLayersLayer 03Layer 04Evidence record: Eval result, Control observation, Test plan, Test report, Design recordEvidence recordEval result+4 moreSame topic elsewhere: no crosswalk topic files this clause yetAs of 2026-09-24 · illustrative, not a claim of conformity From clause to evidenceThe chain from EU AI Act Art. 15(4) to its evidence record, in 6 steps; the text alternative lists them and the facts under the figure state each in full.Clause: EU AI Act (post-Omnibus), Art. 15(4)ClauseEU AI Act · Art. 15Duty holder: ProviderDuty holderProviderApplies from: 2027-12-02 (Deferred)Applies from2027-12-02 · DeferredArtefact: Feedback-loop fairness monitor; retraining-data bias check; agent memory write gate with provenance and rollback to a known-good snapshotArtefactFeedback-loop fairness monitorLayers: Layer 03 Evals & Red Teaming as Evidence and Layer 04 Runtime Controls & ObservabilityLayersLayer 03Layer 04Evidence record: Eval result, Control observation, Test plan, Test report, Design recordEvidence recordEval result · +4 moreSame topic elsewhere: no crosswalk topicfiles this clause yetAs of 2026-09-24illustrative, not a claim of conformity
From clause to evidence Build the artefact, then file every output it produces as a record that names AIGE-OBL-EUAIA-ART15-4. Drawn from chapter 08.
Text alternative
  • Clause: EU AI Act, Art. 15.
  • Duty holder: Provider.
  • Applies from: 2027-12-02, Deferred.
  • Artefact: Feedback-loop fairness monitor.
  • Layers: Layer 03, Layer 04.
  • Evidence record: Eval result, +4 more.
  • Record schemas: Eval result , Control observation , Test plan , Test report , Design record .
  • No crosswalk topic files this clause yet.
When this obligation applies3 dates of AIGE-OBL-EUAIA-ART15-4 on a time axis, against its review date 2026-09-24.2024202620282030As of 2026-09-242027-12-02 · DeferredAnnex III2028-08-02 · High-risk(Annex I)Applies to Annex I embedded(product safety-component)systems2030-08-02 · Later stepDeadline for legacy high-risksystems intended for use bypublic authorities (Art.111(2))2027-12-02 · Deferred · Annex III2028-08-02 · High-risk (Annex I) · Applies to Annex I embedded (product safety-component) systems2030-08-02 · Later step · Deadline for legacy high-risk systems intended for use by public authorities (Art. 111(2))Source: Obligation register(chapter 08)
When this obligation applies3 dates of AIGE-OBL-EUAIA-ART15-4 on a time axis, against its review date 2026-09-24.2027-12-02 · DeferredAnnex III2028-08-02 · High-risk (Annex I)Applies to Annex I embedded(product safety-component)systems2030-08-02 · Later stepDeadline for legacy high-risksystems intended for use by publicauthorities (Art. 111(2))2024202620282030As of 2026-09-242027-12-02 · Deferred · Annex III2028-08-02 · High-risk (Annex I) · Applies to Annex I embedded (product safety-component) systems2030-08-02 · Later step · Deadline for legacy high-risk systems intended for use by public authorities (Art. 111(2))Source: Obligation register (chapter 08)
When this obligation applies The first date in the row's status, then each later step: filled once its date has passed, outlined while it is still ahead of the review date.
Data table: the dates
Dates of AIGE-OBL-EUAIA-ART15-4
DateStepDetail
2027-12-02DeferredAnnex III
2028-08-02High-risk (Annex I)Applies to Annex I embedded (product safety-component) systems
2030-08-02Later stepDeadline for legacy high-risk systems intended for use by public authorities (Art. 111(2))
Id
AIGE-OBL-EUAIA-ART15-4
Instrument
EU AI Act (post-Omnibus) law
Compared side by side
ISO 42001 vs EU AI Act · NIST AI RMF vs EU AI Act
Clause
Art. 15(4)
Duty holder
Provider
Authority
National MSA
Applies from
Deferred · Annex III
Later dates
  • Applies to Annex I embedded (product safety-component) systems
  • Deadline for legacy high-risk systems intended for use by public authorities (Art. 111(2))
System class
High-risk (Annex III) · High-risk (Annex I)

The artefact that evidences it

Feedback-loop fairness monitor; retraining-data bias check; agent memory write gate with provenance and rollback to a known-good snapshot.

What this obligation connects toEU AI Act Art. 15(4) and its 11 relations: 10 patterns, 1 open controls.DirectColour: stack layer of the pattern or control01 Govern-as-Code03 Evals04 Runtime Controls05 AssuranceEU AI Act Art. 15(4)Patterns (10)AI Threat Model · DirectAI Threat ModelAdversarial Red-Team Suite · DirectAdversarial Red-Team SuiteAgent Identity & Scoped Credentials · DirectAgent Identity & Scoped…Claims Substantiation Gate · DirectClaims Substantiation GateDrift & Fairness Monitor · DirectDrift & Fairness MonitorEval Gate in CI · DirectEval Gate in CI+4 more+4 more: Fairness Eval Suite, Kill Switch / Circuit Breaker, Model Artefact Integrity, Runtime GuardrailOpen controls (1)AIGE-CTL-DEPLOY-009 Fairness monitored by group in production · DirectFairness monitored by group…
What this obligation connects toEU AI Act Art. 15(4) and its 11 relations: 10 patterns, 1 open controls.DirectColour: stack layer of the pattern or control01 Govern-as-Code03 Evals04 Runtime Controls05 AssurancePatterns (10)AI Threat Model · DirectAI Threat ModelAdversarial Red-Team Suite · DirectAdversarial Red-Team SuiteAgent Identity & Scoped Credentials · DirectAgent Identity & Scoped…Claims Substantiation Gate · DirectClaims Substantiation GateDrift & Fairness Monitor · DirectDrift & Fairness MonitorEval Gate in CI · DirectEval Gate in CI+4 more+4 more: Fairness Eval Suite, Kill Switch / Circuit Breaker, Model Artefact Integrity, Runtime GuardrailOpen controls (1)AIGE-CTL-DEPLOY-009 Fairness monitored by group in production · DirectFairness monitored by group…EU AI Act Art.15(4)
What this obligation connects to Solid edges are direct relations, dashed ones looser (a case that cites the same article, a framework whose clauses only touch the topic); patterns and controls take the colour of their stack layer. Select a node to open it; the lists below give every relation in full.
Data table: every relation
Relations of AIGE-OBL-EUAIA-ART15-4
FamilyItemRelation
PatternsAI Threat ModelDirect
PatternsAdversarial Red-Team SuiteDirect
PatternsAgent Identity & Scoped CredentialsDirect
PatternsClaims Substantiation GateDirect
PatternsDrift & Fairness MonitorDirect
PatternsEval Gate in CIDirect
PatternsFairness Eval SuiteDirect
PatternsKill Switch / Circuit BreakerDirect
PatternsModel Artefact IntegrityDirect
PatternsRuntime GuardrailDirect
Open controlsAIGE-CTL-DEPLOY-009 Fairness monitored by group in productionDirect

Patterns that build it

Open controls that evidence it

Draft controls in the open control profiles that map to this row: each states a requirement and the evidence it must leave behind.

Source

Chapter 08, section EU AI Act, post-Omnibus, checked against its sources on the review date above.

Machine-readable

Cite this obligation

García Aibar, J. (2026). EU AI Act Art. 15(4): feedback loops in systems that continue to learn (AIGE-OBL-EUAIA-ART15-4). In AI Governance Engineering: The Thesis & Body of Knowledge (v0.5.0). https://doi.org/10.5281/zenodo.22956197. https://aigovernanceengineer.com/obligations/aige-obl-euaia-art15-4. CC BY 4.0

BibTeX

@misc{aige2026obligation,
  author       = {Jorge García Aibar},
  title        = {{EU AI Act Art. 15(4): feedback loops in systems that continue to learn (AIGE-OBL-EUAIA-ART15-4)}},
  howpublished = {In AI Governance Engineering: The Thesis \& Body of Knowledge},
  year         = {2026},
  version      = {0.5.0},
  doi          = {10.5281/zenodo.22956197},
  url          = {https://aigovernanceengineer.com/obligations/aige-obl-euaia-art15-4},
  note         = {Version 0.5.0}
}