If you develop AI systems for structural load assessment, building material property prediction, geotechnical analysis, fire resistance modelling, or BIM-integrated structural evaluation for construction works in the EU, Article 112 of the EU AI Act directly governs your compliance obligations. EU AI Act Art.112 amends Regulation (EU) 305/2011 — the Construction Products Regulation (CPR) — to formally integrate EU AI Act requirements into the construction product conformity framework, extending the high-risk AI pathway to AI-enabled structural assessment tools, material testing prediction platforms, and geotechnical analysis systems used to support the performance declaration and CE marking of construction products in buildings, bridges, tunnels, and civil engineering infrastructure across the EU.

Article 112 is the ninth and final amendment in the Art.104–112 series of EU AI Act sector legislation insertions. Like Art.104 (machinery), Art.105 (agricultural vehicles), Art.106 (two/three-wheel vehicles), Art.107 (motor vehicles), Art.108 (marine equipment), Art.109 (radio equipment), Art.110 (batteries), and Art.111 (pressure equipment), Art.112 operates through the Annex I bridge mechanism: AI systems associated with construction products that qualify as performance assessment tools, structural analysis systems, or safety-relevant testing platforms under Regulation (EU) 305/2011 now sit within EU AI Act Annex I, activating the Art.6(1) high-risk pathway and triggering the full Title III compliance stack.

Art.112 introduces a dual-transition compliance challenge unique within the Art.104–112 series: the CPR itself is being replaced. Regulation (EU) 305/2011 is superseded by a revised Construction Products Regulation under Commission Proposal COM(2022)144 — which introduces mandatory digital product passports for construction products and substantially tightened performance data requirements. Developers building AI systems for CPR compliance must simultaneously prepare for EU AI Act Art.112 obligations under the existing CPR framework and anticipate how the new CPR's digital passport infrastructure will reshape the AI compliance landscape for construction products after the new regulation enters force.

What Regulation (EU) 305/2011 Covers

Regulation (EU) 305/2011 on the harmonisation of the laws relating to the marketing of construction products governs products that are permanently incorporated into construction works — buildings and civil engineering works — and affect the performance of those works with respect to the seven Basic Requirements for Construction Works (BWRs) set out in Annex I of the regulation.

The CPR scope is defined by the product meeting two conditions: it is a construction product (manufactured and placed on the market for permanent incorporation into construction works or parts thereof) and its performance characteristics are governed by a harmonised technical specification (hTS). Harmonised technical specifications take two forms:

Harmonised European Standards (hENs): CEN/CENELEC standards adopted under the CPR mandate, covering products from structural steel and reinforced concrete to thermal insulation and fire doors. When a harmonised standard covers a product, the manufacturer must issue a Declaration of Performance (DoP) and affix CE marking for CPR-covered characteristics.

European Technical Assessments (ETAs): Performance assessments for innovative construction products not covered by harmonised standards, issued by Technical Assessment Bodies (TABs) against European Assessment Documents (EADs).

The seven Basic Requirements for Construction Works in CPR Annex I define the performance dimensions that construction products must support:

CPR conformity assessment for construction products depends on the system of assessment and verification of constancy of performance (AVCP) applicable to the product under its harmonised standard — Systems 1+, 1, 2+, 2, 3, and 4, with higher-numbered systems indicating less stringent third-party involvement.

What Art.112 Actually Changes

Article 112 is a targeted legislative amendment inserting EU AI Act obligations into the CPR conformity framework through the Annex I bridge mechanism. EU AI Act Annex I lists Union harmonisation legislation whose product scope intersects with the high-risk AI classification pathway. Art.112 adds Regulation (EU) 305/2011 to that list.

AI systems that function as performance assessment tools, structural analysis systems, or safety-relevant testing platforms for construction products covered by the CPR become Annex I construction products AI — and follow Art.6(1)'s high-risk classification rule when performing a safety function within or for those construction products.

Three conditions determine whether a construction products AI system becomes high-risk under Art.112:

1. The associated product is a construction product within the scope of Regulation (EU) 305/2011 governed by a harmonised technical specification
2. The AI system performs structural assessment, material performance prediction, or safety-relevant testing for that construction product under the CPR framework
3. The assessment relates to BWR 1 (mechanical resistance and stability) or BWR 2 (safety in case of fire) — the safety-critical basic requirements — or the product requires AVCP System 1+ or 1 conformity assessment involving a Notified Body

Where these conditions are met, developers face dual compliance: the CPR DoP and CE marking process plus the full EU AI Act Title III compliance stack.

The CPR Revision: Art.112's Unique Dual-Transition Challenge

Article 112 creates a compliance situation without parallel in the Art.104–112 series: the regulated product framework it amends is itself being replaced during the same legislative period as the EU AI Act implementation.

Commission Proposal COM(2022)144 for a revised Construction Products Regulation was adopted and the new regulation is entering force in phases. The new CPR introduces several provisions that directly reshape the AI compliance landscape for Art.112:

Mandatory Digital Product Passports for Construction Products: The new CPR requires digital product passports for certain construction product categories — structural steel, concrete, aluminium, timber, and thermal insulation products in the first wave. AI systems that generate, verify, or interpret performance data for inclusion in these digital passports become infrastructure-embedded AI in the construction regulatory ecosystem. Where an AI material testing prediction system supplies compressive strength values, fire resistance classifications, or thermal performance data that feeds directly into a digital product passport used for CPR compliance, that AI system's outputs are regulatory assertions.

Tightened Performance Reliability Requirements: The new CPR strengthens requirements for the accuracy and reliability of declared performance values. AI systems providing declared performance data — predicted compressive strength from AI concrete mix analysis, AI-based thermal resistance estimates — will need to demonstrate performance reliability that satisfies both the new CPR accuracy requirements and EU AI Act Art.9 risk management and Art.15 accuracy obligations.

Third Country Construction Products: The new CPR strengthens market surveillance for construction products from third countries — a significant policy objective given the volume of structural steel, technical ceramics, and insulation products imported from China. AI systems assessing the performance of third-country construction products for EU market entry sit in a regulatory environment under simultaneous pressure from the new CPR's third-country product provisions and AI Act high-risk obligations.

This dual-transition means developers building construction AI systems for 2026 deployment must design for the current CPR (305/2011) framework targeted by Art.112 while building architecture capable of adapting to the new CPR's digital passport and performance reliability regime when it comes into effect.

Definitively High-Risk Construction Products AI Systems

AI Structural Load Assessment Platforms Using Eurocodes

AI systems performing structural load assessment and stability analysis under the Eurocodes — the harmonised suite of structural design standards (EN 1990 through EN 1999) — for construction products assessed under CPR BWR 1 are definitively high-risk. The Eurocodes govern how structural loads are calculated and how structural members must be sized to resist those loads:

AI platforms that automate Eurocode-based structural design checks — taking material property inputs, applied loads, and geometric parameters and producing code-compliant structural solutions — are definitively high-risk when used to support CPR performance assessment for construction products subject to System 1+ or 1 AVCP (involving Notified Bodies). AI errors in load calculation, partial factor application, or material property characterisation can propagate into construction product specifications and structural designs where the consequences of failure include collapse of buildings or infrastructure.

Providers operating in this space include Nemetschek Group (Allplan, Vectorworks), Autodesk (Revit Structural, Robot Structural Analysis), Bentley Systems (STAAD.Pro, RAM Structural System), Dlubal Software (RFEM, RSTAB), and SCIA (SCIA Engineer). AI enhancements to these platforms — AI-assisted load combinations, AI section optimisation, AI structural health monitoring integration — require Art.112 compliance assessment.

AI Fire Resistance Assessment Systems

AI systems modelling fire resistance behaviour of construction products — predicting fire resistance periods, smoke spread characteristics, or load-bearing capacity under fire conditions per CPR BWR 2 — are definitively high-risk. Fire resistance classification is a mandatory declared performance characteristic for structural elements (columns, beams, slabs, walls) and building envelope products (fire doors, fire-rated glazing, cavity barriers) in most CPR harmonised standards.

AI fire resistance models trained on furnace test data to predict fire resistance ratings from product geometry and material properties, AI computational fluid dynamics models predicting smoke spread in buildings for evacuation design, and AI structural fire engineering systems assessing load-bearing capacity reduction under EN 1993-1-2 (steel structures in fire) or EN 1992-1-2 (concrete structures in fire) are definitively high-risk where they generate or support CPR declared performance values that go into the DoP and CE marking.

AI Geotechnical Analysis Systems for Foundations

AI systems performing geotechnical analysis under EN 1997 to assess ground conditions, predict bearing capacity, calculate settlement, and design foundations for construction works are definitively high-risk when their outputs feed into structural product assessments or building designs under CPR BWR 1. Foundation failures are among the highest-consequence structural failure modes — building settlements and tilting can render structures uninhabitable or lead to collapse.

AI ground condition characterisation from borehole and CPT data, AI bearing capacity prediction for pile foundations, AI slope stability assessment for retaining walls and embankments, and AI settlement prediction for buildings on variable ground are definitively high-risk AI systems where they provide performance inputs used in CPR DoP preparation or structural design certification.

Conditionally High-Risk Construction Products AI Systems

AI Building Material Property Prediction

AI systems predicting material performance characteristics from mix design, manufacturing process data, or non-destructive measurement — AI concrete compressive strength prediction from mix proportions and curing parameters, AI prediction of thermal resistance from insulation product density and thickness, AI prediction of acoustic performance from material structure — are conditionally high-risk.

The high-risk condition is whether the AI prediction output is used directly as a declared performance value in a CPR Declaration of Performance. Where an AI concrete strength prediction model is used to generate the characteristic compressive strength fck that appears in the DoP for a structural concrete product under EN 206 — without independent physical testing verification — the AI system is performing a regulatory assertion that directly determines whether the product may be CE marked for structural applications. In that deployment context, it is definitively high-risk. Where AI prediction is used to optimise the production process but all DoP values are still determined by physical test results, the AI system is conditionally high-risk.

AI BIM-Integrated Structural Analysis

AI systems integrated into Building Information Modelling (BIM) platforms that perform structural analysis, detect structural design conflicts, or automate compliance checks are conditionally high-risk. BIM platforms — Revit, Archicad, Vectorworks, BIM 360 — increasingly incorporate AI to automate repetitive analysis tasks: AI clash detection for structural and MEP coordination, AI regulatory compliance checking against national building codes and Eurocodes, AI structural optimisation suggesting design modifications to reduce material use while maintaining code compliance.

The high-risk condition is whether the AI BIM analysis output directly determines a CPR performance declaration or replaces a licensed structural engineer's assessment for code-compliance purposes. Where AI BIM structural checks are reviewed and approved by a qualified structural engineer before submission for planning permission or building regulation approval, the system is conditionally high-risk. Where AI BIM output is used to auto-generate declarations without engineer review, it crosses into the definitively high-risk category.

AI Acoustic and Thermal Performance Prediction

AI systems predicting acoustic performance (airborne sound insulation Rw, impact sound insulation Ln,w) and thermal performance (thermal transmittance U-value, thermal resistance R-value) of building products from design parameters are conditionally high-risk. These predictions support CPR BWR 5 and BWR 6 performance declarations. The high-risk condition is whether AI prediction replaces physical laboratory testing as the basis for declared performance values under the applicable harmonised standard.

Provider-Deployer Split in Construction Products AI

Construction AI compliance responsibility is distributed across structural software vendors, construction engineering consultancies, material manufacturers, and certification bodies.

Structural Software Vendors as Providers: Companies supplying AI-enhanced structural analysis and design software — Nemetschek Group, Autodesk, Bentley Systems, Dlubal, SCIA, Graitec — are EU AI Act providers when their AI structural assessment and load analysis functions perform safety determinations for CPR-regulated construction products. They must maintain technical documentation under Art.11, implement QMS under Art.17, and ensure AI structural analysis features undergo conformity assessment before deployment in regulated construction product assessment workflows.

Engineering Consultancies as Both Providers and Deployers: Global engineering consultancies — Arup, AECOM, WSP, Arcadis, Jacobs — occupy a dual role. When they develop proprietary AI tools for structural assessment, seismic analysis, or geotechnical modelling, they are providers. When they deploy AI tools developed by structural software vendors as part of project services, they are deployers with Art.26 obligations: verifying the AI tool's conformity documentation, maintaining logs of AI analysis outputs, and reporting anomalous AI performance to the provider.

Construction Product Manufacturers as Deployers: Manufacturers of structural steel (ArcelorMittal, Tata Steel), concrete products (HeidelbergMaterials, Holcim, CEMEX), and technical insulation (Rockwool, Saint-Gobain, Knauf) that deploy AI systems to predict or support CPR declared performance values are EU AI Act deployers. Their Art.26 obligations include: verifying AI tools used in performance testing programmes carry appropriate conformity documentation, maintaining logs of AI assessment outputs used to support DoP preparation, and implementing human oversight mechanisms before AI predictions are used as declared performance values.

Notified Bodies and TABs in the AI Ecosystem: Notified Bodies and Technical Assessment Bodies that use AI tools to assist in CPR conformity assessment — AI-assisted review of product test data, AI-automated analysis of ETA applications — occupy the same dual-role compliance position as notified bodies in the PED context (Art.111). The NB using AI in its conformity assessment work is a deployer subject to EU AI Act deployer obligations for the AI tools it uses.

CLOUD Act Intersection for Construction Products AI

Construction structural analysis data, BIM models, and material testing datasets are increasingly stored in US-provider cloud infrastructure, creating CLOUD Act exposure for EU construction sector operators.

Structural Design Data in US Cloud: Major BIM and structural analysis platforms are US-headquartered companies operating cloud infrastructure subject to US jurisdiction. Autodesk (BIM 360, ACC), Bentley Systems (iTwin), and Trimble (Tekla) operate cloud-based construction collaboration platforms where project structural models, load calculations, and CPR conformity documentation are stored. Under the US CLOUD Act, US law enforcement can compel these providers to produce stored data regardless of server location.

Proprietary AI Training Data from Infrastructure Projects: AI models trained on structural design data from EU infrastructure projects — bridges, tunnels, high-rise buildings — incorporate proprietary engineering methodologies and performance data from projects with potential national security relevance. AI models trained on datasets from critical infrastructure design projects should not reside in CLOUD Act-exposed environments. The combination of structural weaknesses data, geotechnical ground condition models, and seismic vulnerability assessments for EU infrastructure constitutes operationally sensitive information.

Critical Infrastructure Design Sensitivity: Structural analysis data for critical infrastructure — rail tunnels, motorway bridges, airport terminals, nuclear facility structural containment — contains information relevant to national infrastructure security. BIM models containing structural schematics, load-bearing element specifications, and identified structural vulnerabilities are sensitive national infrastructure data. Storing this data in US-jurisdiction cloud environments creates data sovereignty risks with national security dimensions.

EU Sovereign Deployment for Infrastructure AI: Engineering firms working on EU critical infrastructure projects, national transportation authorities, and public sector construction clients should specify EU-sovereign deployment for structural AI platforms handling infrastructure-sensitive data. For operators subject to NIS2's critical infrastructure provisions, EU-jurisdiction data residency for structural AI processing may be a compliance requirement rather than a commercial choice.

Technical Standards Intersection

Construction products AI systems must be evaluated against EU AI Act technical documentation requirements and applicable harmonised standards and Eurocode provisions:

AI systems providing performance assessments or design inputs for construction products governed by these standards must maintain EU AI Act Art.11 and Annex IV technical documentation cross-referencing the applicable harmonised standard and mapping the AI assessment methodology to the standard's performance characterisation and acceptance criteria framework.

Python Construction Products AI Compliance Tracker

from dataclasses import dataclass, field
from typing import Literal

AVCPSystem = Literal["1+", "1", "2+", "2", "3", "4"]
AIFunction = Literal[
    "structural_load_assessment", "fire_resistance_modelling",
    "geotechnical_analysis", "material_property_prediction",
    "bim_structural_analysis", "acoustic_performance", "thermal_performance"
]
BWR = Literal["bwr1_structural", "bwr2_fire", "bwr3_health", "bwr4_safety",
              "bwr5_noise", "bwr6_energy", "bwr7_sustainability"]
CloudProvider = Literal["aws", "azure", "gcp", "eu_sovereign", "on_premise"]

@dataclass
class ConstructionAISystem:
    name: str
    avcp_system: AVCPSystem
    ai_function: AIFunction
    primary_bwr: BWR
    cloud_provider: CloudProvider
    output_used_in_dop: bool
    notified_body_involved: bool
    replaces_physical_testing: bool = False

class ConstructionAIComplianceTracker:
    def __init__(self):
        self.systems: list[ConstructionAISystem] = []

    def add(self, s: ConstructionAISystem) -> None:
        self.systems.append(s)

    def classify(self, s: ConstructionAISystem) -> str:
        # Definitively high-risk: safety-critical BWR with notified body + DoP output
        if s.primary_bwr in ("bwr1_structural", "bwr2_fire") and \
                s.notified_body_involved and s.output_used_in_dop:
            return "DEFINITIVELY_HIGH_RISK"
        # Definitively high-risk: replaces physical testing for DoP
        if s.replaces_physical_testing and s.output_used_in_dop and \
                s.avcp_system in ("1+", "1"):
            return "DEFINITIVELY_HIGH_RISK"
        # Definitively high-risk: structural/fire AI in AVCP 1+ or 1
        if s.ai_function in ("structural_load_assessment", "fire_resistance_modelling",
                              "geotechnical_analysis") and \
                s.avcp_system in ("1+", "1"):
            return "DEFINITIVELY_HIGH_RISK"
        # Conditionally high-risk: performance-related but lower AVCP or indirect
        if s.primary_bwr in ("bwr1_structural", "bwr2_fire"):
            return "CONDITIONALLY_HIGH_RISK"
        if s.ai_function == "bim_structural_analysis":
            return "CONDITIONALLY_HIGH_RISK"
        if s.output_used_in_dop:
            return "CONDITIONALLY_HIGH_RISK"
        return "NOT_HIGH_RISK"

    def cloud_act_exposure(self, s: ConstructionAISystem) -> bool:
        return s.cloud_provider in ("aws", "azure", "gcp")

    def cpr_revision_risk(self, s: ConstructionAISystem) -> bool:
        # Systems that will be affected by new CPR digital passport requirements
        return s.output_used_in_dop or s.replaces_physical_testing

    def compliance_summary(self) -> dict:
        definitely = [s for s in self.systems if self.classify(s) == "DEFINITIVELY_HIGH_RISK"]
        conditional = [s for s in self.systems if self.classify(s) == "CONDITIONALLY_HIGH_RISK"]
        not_high = [s for s in self.systems if self.classify(s) == "NOT_HIGH_RISK"]
        cloud_risk = [s for s in self.systems if self.cloud_act_exposure(s)]
        cpr_rev = [s for s in self.systems if self.cpr_revision_risk(s)]
        return {
            "total": len(self.systems),
            "definitely_high_risk": len(definitely),
            "conditionally_high_risk": len(conditional),
            "not_high_risk": len(not_high),
            "cloud_act_exposure": len(cloud_risk),
            "cpr_revision_affected": len(cpr_rev),
            "dual_compliance_required": len(definitely) + len(conditional),
        }

tracker = ConstructionAIComplianceTracker()
tracker.add(ConstructionAISystem(
    name="AI Eurocode structural load assessment — structural steel AVCP 1+",
    avcp_system="1+", ai_function="structural_load_assessment",
    primary_bwr="bwr1_structural", cloud_provider="azure",
    output_used_in_dop=True, notified_body_involved=True,
    replaces_physical_testing=False))
tracker.add(ConstructionAISystem(
    name="AI fire resistance modelling — fire door EN 16034 AVCP 1",
    avcp_system="1", ai_function="fire_resistance_modelling",
    primary_bwr="bwr2_fire", cloud_provider="aws",
    output_used_in_dop=True, notified_body_involved=True,
    replaces_physical_testing=False))
tracker.add(ConstructionAISystem(
    name="AI concrete compressive strength prediction replaces testing",
    avcp_system="1+", ai_function="material_property_prediction",
    primary_bwr="bwr1_structural", cloud_provider="eu_sovereign",
    output_used_in_dop=True, notified_body_involved=True,
    replaces_physical_testing=True))
tracker.add(ConstructionAISystem(
    name="AI BIM structural clash detection and compliance check",
    avcp_system="2+", ai_function="bim_structural_analysis",
    primary_bwr="bwr1_structural", cloud_provider="aws",
    output_used_in_dop=False, notified_body_involved=False,
    replaces_physical_testing=False))
tracker.add(ConstructionAISystem(
    name="AI U-value thermal performance prediction — insulation AVCP 3",
    avcp_system="3", ai_function="thermal_performance",
    primary_bwr="bwr6_energy", cloud_provider="azure",
    output_used_in_dop=True, notified_body_involved=False,
    replaces_physical_testing=False))

summary = tracker.compliance_summary()
# {"total": 5, "definitely_high_risk": 3, "conditionally_high_risk": 1,
#  "not_high_risk": 1, "cloud_act_exposure": 3, "cpr_revision_affected": 4,
#  "dual_compliance_required": 4}

The Art.104–112 Amendment Series: Complete

Article 112 is the ninth and final amendment in the Art.104–112 series. Together, these articles form a comprehensive integration of EU AI Act obligations into the major EU product safety regulation frameworks, creating a unified high-risk AI pathway across the full spectrum of manufactured and constructed products used in the EU economy:

The completion of the Art.104–112 series with Art.112 closes the EU AI Act's systematic integration of high-risk AI obligations into EU product safety law. Every major sector of the EU single market for manufactured products now has an explicit AI Act nexus: from the factory floor (machinery, Art.104) through transportation (vehicles, Art.105–107), maritime (Art.108), communications (Art.109), energy storage (Art.110), industrial infrastructure (Art.111), and the built environment (Art.112). Developers building AI for any of these sectors face dual compliance from August 2026.

25-Item Construction Products AI Compliance Checklist (Art.112)

Product and System Scoping

1. Confirm the associated construction product falls within Regulation (EU) 305/2011 scope and is covered by a harmonised technical specification (hEN or ETA)
2. Determine the AVCP system applicable to the product under its harmonised standard — Systems 1+ and 1 trigger the Art.112 high-risk threshold via Notified Body involvement
3. Identify the primary Basic Requirements for Construction Works (BWRs) the AI system's outputs relate to — BWR 1 (structural) and BWR 2 (fire) are the safety-critical thresholds
4. Map all AI functions: structural load assessment, fire resistance modelling, geotechnical analysis, material property prediction, BIM structural analysis, acoustic/thermal performance
5. Determine whether AI outputs are used as declared performance values in a CPR Declaration of Performance — if yes, the Art.112 high-risk pathway is activated

High-Risk Classification 6. Apply Art.6(1) test: does the AI perform a safety function for a CPR construction product subject to AVCP System 1+ or 1? 7. Assess whether AI material property prediction replaces physical laboratory testing as the basis for CPR DoP declared performance values 8. Evaluate AI structural load assessment tools: do AI outputs directly determine CPR performance declarations without independent engineering review? 9. For AI BIM structural analysis: determine whether AI output is reviewed by a licensed structural engineer before regulatory submission or is used to auto-generate compliance documentation 10. Document classification decision for each AI system with regulatory basis under both CPR and the EU AI Act

Dual Conformity Assessment 11. Plan CPR conformity path for AI systems supporting new construction product assessments and identify applicable AVCP System and Notified Body 12. For AI systems supporting in-use structural assessment: identify the applicable conformity assessment path under EU AI Act Title III 13. Establish EU AI Act technical documentation (Art.11, Annex IV) for each high-risk construction AI system cross-referencing applicable Eurocode and harmonised standard methodology 14. Implement AI risk management system (Art.9) with construction-specific failure mode analysis: structural overload, foundation failure, fire progression, seismic collapse 15. Define human oversight mechanisms (Art.14) for AI structural assessments and AI material performance predictions used in CPR DoP preparation

CPR Revision Readiness 16. Assess which AI systems will be affected by the new CPR's digital product passport requirements — identify AI outputs that will feed into mandatory construction product passports 17. Review whether AI material performance prediction reliability meets the new CPR's tightened declared performance accuracy requirements 18. Evaluate AI systems for third-country construction product assessment — the new CPR's enhanced market surveillance for imported products may create additional AI compliance touchpoints 19. Map AI system architecture for adaptability to new CPR data format and digital passport interface requirements before the new CPR enters force 20. Document CPR transition plan: current 305/2011 compliance + new CPR readiness roadmap

Data, Cloud, and Post-Market 21. Inventory structural analysis and BIM data flows to cloud platforms: identify CLOUD Act exposure for project structural models, load calculations, and CPR conformity documentation 22. Assess critical infrastructure data sensitivity: structural analysis data for bridges, tunnels, and public buildings may require EU-sovereign storage under NIS2 or national critical infrastructure rules 23. Implement AI performance monitoring under Art.72: track structural AI output accuracy against project outcomes, construction defects attributable to AI design inputs, and material AI prediction error rates against physical test results 24. Establish AI anomaly reporting chain: AI structural assessment anomalies, AI material prediction failures, and AI fire modelling divergences must flow from operator to AI system provider under Art.26(5) 25. Register high-risk construction products AI systems in EU AI Act database (Art.71) before August 2026 application deadline

See Also

• EU AI Act Annex III Point 1: Biometric Identification, Categorisation and Emotion Recognition — Prohibited vs High-Risk Classification
• EU AI Act Art.111: Amendment to Directive 2014/68/EU — AI Structural Integrity Monitoring, NDT Analysis and Pressure Vessel Inspection High-Risk Classification
• EU AI Act Art.110: Amendment to Directive 2006/66/EC — AI Battery Management, State-of-Health Estimation and EV BMS High-Risk Classification
• EU AI Act Art.109: Amendment to Directive 2014/53/EU — Software-Defined Radio, Cognitive Radio and AI Spectrum Management High-Risk Classification
• EU AI Act Art.108: Amendment to Directive 2014/90/EU — Marine Equipment ECDIS and Autopilot AI
• EU AI Act Art.107: Amendment to Regulation (EU) 2019/2144 — ALKS, ADAS and Motor Vehicle AI High-Risk Classification
• EU AI Act Art.104: Amendment to Directive 2006/42/EC — Machinery Safety AI High-Risk Classification