Cyber Resilience Act: The New Product-Security Baseline

Scope, legal roles, secure design, conformity, reporting, and lifecycle assurance for European SMEs

A technical and legal analysis of the EU Cyber Resilience Act, explaining product scope, economic-operator roles, essential cybersecurity requirements, conformity assessment, component dependencies, vulnerability reporting, transition, and practical assurance governance.
cybersecurity
regulation and compliance
🇬🇧
Author
Affiliation

Antonio Montano

4M4

Published

August 3, 2026

Modified

August 4, 2026

Abstract

Regulation (EU) 2024/2847—the Cyber Resilience Act—changes European cybersecurity law by applying mandatory, horizontal product-security requirements to connected hardware, software, and qualifying remote data-processing solutions placed on the Union market. Its central innovation is not a particular security control. It is the conversion of cybersecurity from a fragmented contractual or sectoral concern into an enforceable product-lifecycle obligation.

This article reconstructs the Regulation from first principles. It distinguishes products from services and integrated systems; assigns manufacturer, importer, distributor, integrator, and user roles; derives the risk-based security baseline in Annex I; and explains how product classification determines the permitted conformity-assessment route. It examines component due diligence, software bills of materials, free and open-source software, substantial modification, security updates, support periods, Article 14 reporting, market surveillance, administrative penalties, and interaction with product-liability law.

The analysis also develops a technical assurance model for procurement and lifecycle governance. That model treats the regulated object as a controlled product state linked to legal roles, requirements, evidence, dependencies, support capability, vulnerability processes, and change decisions. It distinguishes legal conformity from internal assurance and from actual resistance to attack.

The implementation environment remains incomplete as of 2 August 2026. Commission guidance is non-binding; harmonised standards and notified-body capacity are still developing; and ENISA’s Single Reporting Platform is scheduled to become operational when Article 14 applies on 11 September 2026. These limitations do not suspend the CRA. They make explicit assumptions, evidence alignment, monitoring, and controlled reassessment essential to defensible compliance.

A technical and legal analysis of the EU Cyber Resilience Act, explaining product scope, economic-operator roles, essential cybersecurity requirements, conformity assessment, component dependencies, vulnerability reporting, transition, and practical assurance governance.

Why the CRA changes the EU cybersecurity architecture

The Cyber Resilience Act (CRA) establishes a horizontal product-security regime for software and hardware made available on the European Union market. It regulates four connected subjects: market access, product-security properties, manufacturer vulnerability-handling processes, and public enforcement.1

The Regulation addresses a structural gap in the earlier cybersecurity architecture. NIS2 imposes governance, risk-management, supply-chain, and incident-reporting duties on covered entities.2 DORA establishes operational-resilience and ICT-risk obligations for the financial sector.3 Neither instrument is a general market-access law requiring every connected software or hardware product to satisfy a common security baseline before placement on the Union market.

The CRA adopts the logic of Union product legislation. A manufacturer defines and assesses a product, demonstrates conformity, issues an EU declaration of conformity, and affixes CE marking. Importers and distributors act as supply-chain gatekeepers. Market-surveillance authorities can inspect evidence, require correction, restrict supply, withdraw products, or order recalls.

Cybersecurity consequently becomes a condition of lawful market placement rather than merely:

  • a contractual service level;
  • a voluntary certification claim;
  • a feature offered in a premium product tier;
  • an internal development policy; or
  • a risk accepted entirely by the customer.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart LR
    MFR([Manufacturer]):::entity
    PRODUCT[Product with digital elements]:::product
    USER([Customer or operator]):::entity

    CRA[/Cyber Resilience Act<br/>product design, conformity<br/>support and reporting/]:::overlay
    NIS2[/NIS2<br/>entity governance and<br/>risk management/]:::overlay
    DORA[/DORA<br/>financial-sector<br/>operational resilience/]:::overlay
    SECTOR[/Sectoral product and<br/>safety legislation/]:::overlay
    AI[/AI Act<br/>high-risk AI requirements/]:::overlay
    LIABILITY[/Product-liability law/]:::overlay

    MFR --> PRODUCT
    PRODUCT --> USER
    CRA ==> PRODUCT
    NIS2 ==> USER
    DORA ==> USER
    SECTOR ==> PRODUCT
    AI ==> PRODUCT
    LIABILITY ==> PRODUCT

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
Figure 1: The CRA adds a horizontal product-lifecycle layer to entity-level, sectoral, safety, and liability regimes.

The layers in Figure 1 can apply simultaneously. A connected machine can be subject to both the CRA and the Machinery Regulation.4 A product incorporating a high-risk AI system can engage both the CRA and the AI Act, with specific coordination rules for cybersecurity requirements and conformity assessment.5 A defective or inadequately updated product can also generate civil-liability questions independently of CRA market-surveillance enforcement.6

This architecture produces an important distinction:

The CRA regulates the cybersecurity of the product and the manufacturer processes that sustain it. Entity-level laws regulate how organisations govern and use technology.

The distinction is not absolute. Product vulnerabilities can generate organisational incidents, and organisational security failures can compromise product build or update systems. The legal objects and compliance evidence nevertheless remain different.

The CRA entered into force on 10 December 2024. Its conformity-assessment-body provisions have applied since 11 June 2026; Article 14 reporting applies from 11 September 2026; and the Regulation applies generally from 11 December 2027.7 The staggered timetable means that reporting readiness must precede completion of the full product-conformity programme.

Scope: which products, software, components, and remote services are regulated

The scope inquiry begins with three cumulative questions:

  1. Is there a software or hardware product, including a separately marketed component or qualifying remote data-processing solution?
  2. Does its intended purpose or reasonably foreseeable use include a direct or indirect logical or physical data connection to a device or network?
  3. Is it made available on the Union market in the course of commercial activity?

A positive answer to all three establishes presumptive scope, subject to the Regulation’s exclusions and sector-specific limitations.8

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    CANDIDATE[Software, hardware<br/>separately marketed component<br/>or remote processing]:::product
    PRODUCT{Product with digital elements<br/>or part of one?}:::test
    CONNECTION{Intended purpose or foreseeable use<br/>includes direct or indirect<br/>logical or physical connection?}:::test
    MARKET{Made available on the Union market<br/>during commercial activity?}:::test
    EXCLUSION{Article 2 exclusion or<br/>sector-specific limitation?}:::test

    INSCOPE[CRA applies]:::product
    SERVICE[Outside the CRA product definition<br/>assess other legislation]:::context
    INTERNAL[No market supply on these facts<br/>reassess if distribution changes]:::context
    EXCLUDED[CRA excluded or limited<br/>to the defined extent]:::context

    CANDIDATE --> PRODUCT
    PRODUCT -- Yes --> CONNECTION
    PRODUCT -- No --> SERVICE
    CONNECTION -- Yes --> MARKET
    CONNECTION -- No --> SERVICE
    MARKET -- Yes --> EXCLUSION
    MARKET -- No --> INTERNAL
    EXCLUSION -- No --> INSCOPE
    EXCLUSION -- Yes --> EXCLUDED

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef context fill:#ffffff,stroke:#7a7a7a,stroke-width:1.5px,stroke-dasharray:5 5,color:#404040;
Figure 2: CRA scope follows a sequence of product, connection, market, and exclusion tests.

Software and components are products

Standalone software can be a product with digital elements. The delivery mechanism is not decisive: an application can be downloaded, supplied through a repository, embedded in hardware, or delivered through another electronic channel.

Source code that is merely stored or developed internally is not necessarily a marketed product. The analysis changes when a controlled version is supplied for distribution or use in commercial activity. Public archives of historical software can remain available, but unsupported versions must be clearly identified as such where the CRA conditions are engaged.9

Hardware and software components are independently regulated where they are placed on the market separately. A library, processor, firmware module, network interface, operating system, or secure element can therefore be both:

  • a product in its own right; and
  • a dependency inside a different downstream product.

The component’s own conformity does not establish the conformity of the downstream integration.

Connectivity is broader than Internet access

A direct physical connection can use electrical, optical, mechanical, wired, or radio interfaces. A logical connection can use an API, file exchange, network socket, inter-process mechanism, or another software interface.

An indirect connection exists where the product communicates through a larger system. A component need not possess an independent Internet interface to fall within scope. Firmware that exchanges data through a host device, for example, can be indirectly connected.

Remote processing enters the product boundary only under specific conditions

A remote data-processing solution is software processing performed at a distance, designed or developed by or under the responsibility of the manufacturer, whose absence would prevent the local product from performing one of its functions.10

The definition excludes two common overstatements:

  • not every cloud service used by the manufacturer becomes part of the product; and
  • a manufacturer cannot exclude essential manufacturer-controlled backend functionality merely because it runs on third-party infrastructure.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    LOCAL[Local software or hardware]:::product
    REMOTE[/Remote processing candidate/]:::overlay
    RESPONSIBILITY{Designed by or under<br/>manufacturer responsibility?}:::test
    NECESSARY{Would absence prevent<br/>a product function?}:::test
    BOUNDARY[[Regulated product boundary<br/>local product plus qualifying<br/>remote processing]]:::system
    EXTERNAL[Independent cloud or SaaS service]:::context
    RISK[External dependency and<br/>integration risk assessment]:::product

    LOCAL --> BOUNDARY
    REMOTE --> RESPONSIBILITY
    RESPONSIBILITY -- Yes --> NECESSARY
    RESPONSIBILITY -- No --> EXTERNAL
    NECESSARY -- Yes --> BOUNDARY
    NECESSARY -- No --> EXTERNAL
    EXTERNAL -. dependency .-> RISK
    RISK -. informs .-> BOUNDARY

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef system fill:#f1f3f5,stroke:#5c6770,stroke-width:1.5px,color:#252a2e;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef context fill:#ffffff,stroke:#7a7a7a,stroke-width:1.5px,stroke-dasharray:5 5,color:#404040;
Figure 3: The product boundary follows functional necessity and manufacturer responsibility, not the location of computation.

The 2026 Commission guidance applies this analysis to mobile applications, connected consumer products, industrial systems, and telecommunications architectures. It remains non-binding and requires case-specific application.11

Complex systems are not automatically one product

A factory, building-management installation, telecommunications network, or enterprise platform can contain many regulated products without becoming a single CRA product merely through technical integration.

The relevant questions are:

  • what functional whole is offered;
  • who controls its design;
  • under whose name it is marketed;
  • whether the complete system is supplied as a product;
  • which parts are independently placed on the market; and
  • whether later integration or modification creates a new product state.

Procurement descriptions such as “solution,” “platform,” and “system” are not legal conclusions.

Exclusions must be applied narrowly

Article 2 excludes, among other things, products governed by specified medical-device, in-vitro-diagnostic, automotive, aviation, and marine-equipment regimes; products developed or modified exclusively for national-security or defence purposes; products specifically designed to process classified information; and identical spare parts made to the same specifications.12

The Commission can also limit or exclude CRA application for products covered by other Union rules where those rules address the same risks at an equivalent or higher level. An exclusion should therefore be documented against the exact product and sectoral provision. The existence of another regulatory regime is not, by itself, a general exemption.

The security baseline: risk assessment, secure design, defaults, and updates

The CRA does not prescribe one identical control catalogue for every product. It establishes objective-oriented requirements whose implementation is informed by the product’s cybersecurity risk assessment.

Article 13 requires the manufacturer to assess risks and use the result throughout planning, design, development, production, delivery, and maintenance. The assessment must address intended purpose, reasonably foreseeable use, operating conditions, protected assets, expected use time, applicable Annex I requirements, and the manner in which they are implemented.14

A requirement can be treated as inapplicable only with a clear justification in the technical documentation. “Risk based” therefore does not mean optional. It means that the manufacturer must explain the relationship between the product context, identified risks, selected controls, verification evidence, and residual risk.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    MFR([Manufacturer]):::entity
    CONTEXT[Purpose, use, environment<br/>assets, interfaces and lifetime]:::product
    RISK[Cybersecurity risk assessment]:::product
    CONTROLS[Applicable Annex I controls]:::product
    BUILD[Design, development<br/>production and verification]:::product
    RELEASE[Controlled market release]:::product
    MONITOR[Support, monitoring<br/>vulnerability handling and updates]:::product
    CHANGE{New vulnerability, incident<br/>dependency or product change?}:::test
    CORRECT{{Correct, restrict<br/>withdraw or recall}}:::enforcement

    MFR --> CONTEXT
    CONTEXT --> RISK
    RISK --> CONTROLS
    CONTROLS --> BUILD
    BUILD --> RELEASE
    RELEASE --> MONITOR
    MONITOR --> CHANGE
    CHANGE -- Reassessment required --> RISK
    CHANGE -- Non-conformity unresolved --> CORRECT
    CHANGE -- No material change --> MONITOR

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 5: The CRA security baseline is a lifecycle feedback process rather than a one-time pre-market checklist.

Annex I combines product properties and manufacturer processes

Annex I, Part I concerns the security properties of the product. Its risk-dependent objectives include:

  • an appropriate level of cybersecurity;
  • release without known exploitable vulnerabilities where applicable;
  • secure-by-default configuration;
  • update capability, including automatic updates where applicable;
  • access control;
  • confidentiality and integrity;
  • data minimisation;
  • availability and resilience;
  • reduction of harmful effects on connected systems;
  • attack-surface limitation;
  • exploitation mitigation;
  • security-relevant recording and monitoring; and
  • secure removal or transfer of user data.

Part II concerns vulnerability handling. It requires component and vulnerability records, an SBOM covering at least top-level software dependencies, remediation, regular testing and review, coordinated disclosure, secure update distribution, and prompt communication of security updates.15

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TB
    RISK[Cybersecurity risk assessment]:::product
    PARTI[Annex I Part I<br/>product-security properties]:::product
    PARTII[Annex I Part II<br/>vulnerability-handling processes]:::product

    DEFAULTS[Secure defaults<br/>access control and data protection]:::product
    RESILIENCE[Availability, attack-surface reduction<br/>monitoring and exploitation mitigation]:::product
    UPDATECAP[Secure update capability]:::product

    INVENTORY[Components, vulnerabilities<br/>and software bill of materials]:::product
    PROCESS[Testing, disclosure<br/>remediation and advisories]:::product
    DISTRIBUTION[Secure update distribution]:::product

    EVIDENCE[Technical documentation<br/>and support-period evidence]:::product

    RISK --> PARTI
    RISK --> PARTII
    PARTI --> DEFAULTS
    PARTI --> RESILIENCE
    PARTI --> UPDATECAP
    PARTII --> INVENTORY
    PARTII --> PROCESS
    PARTII --> DISTRIBUTION
    DEFAULTS --> EVIDENCE
    RESILIENCE --> EVIDENCE
    UPDATECAP --> EVIDENCE
    INVENTORY --> EVIDENCE
    PROCESS --> EVIDENCE
    DISTRIBUTION --> EVIDENCE

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
Figure 6: Annex I connects risk-selected product properties with continuing vulnerability-handling processes.

Secure by default concerns the delivered state

The secure-by-default requirement is evaluated against the configuration supplied by the manufacturer. Depending on the product and risk, that can require:

  • unnecessary interfaces and services to be disabled;
  • access control and validation to be enabled;
  • shared or predictable credentials to be avoided;
  • debugging and maintenance functions to be restricted;
  • insecure legacy modes to require deliberate activation;
  • security logging to be appropriately configured; and
  • automatic security updating to be enabled where applicable.

The Regulation can permit an explicitly agreed alternative for a tailor-made product supplied to a business user. Ordinary configuration options in a mass-market product do not create that exception.16

“Known exploitable” is a release decision, not a promise of perfection

The CRA does not require proof that a non-trivial product contains no unknown defect. It requires the manufacturer to evaluate known vulnerabilities in their operational context and to avoid market placement with a known exploitable vulnerability where the requirement applies.

The analysis should consider attacker access, preconditions, affected versions, exploit feasibility, available mitigations, severity, connected-system effects, and whether release itself supplies an urgent correction. The reasoning and evidence should be preserved.

Update capability is part of the architecture

Security updates are not merely an optional support service. The product must be designed so that vulnerabilities can be addressed through updates. Where automatic installation is applicable, it should be enabled by default, operate within an appropriate timeframe, permit understandable opt-out and temporary postponement, and notify users.

A credible update system can require:

  • authenticated packages;
  • integrity-protected metadata;
  • signing-key governance;
  • version and compatibility checks;
  • rollback protection;
  • staged deployment;
  • failure-safe installation;
  • recovery capability; and
  • monitoring of update success.

Security updates must normally be made available without delay and free of charge. They should be separable from functionality updates where technically feasible. A user’s refusal to install a properly supplied update does not cure a defective update mechanism or inadequate user information.

Support periods must follow expected use

The support period must reflect the time the product is expected to be used, considering reasonable user expectations, intended purpose, comparable products, operating-environment availability, applicable law, and support for core third-party components.

The period is ordinarily at least five years. A shorter period is permitted where expected use is genuinely shorter. Products expected to remain in use longer require a correspondingly longer analysis.17

The manufacturer must disclose the support end date, including at least month and year, at purchase. Each security update issued during support must remain available for at least ten years after issue or for the remainder of the support period, whichever is longer.

A five-year label is therefore not a universal safe harbour. A five-year support commitment for a product embedded in twenty-year industrial equipment can expose a mismatch that the final-product manufacturer or integrator must resolve.

Classification and conformity assessment

The CRA separates substantive security from procedural assurance. The cybersecurity risk assessment determines what security treatment the product needs. Classification determines which procedure the manufacturer may use to demonstrate conformity.

A product outside Annexes III and IV is often described by the Commission as belonging to the default category. It can use module A internal control. Products whose core functionality matches Annex III are important class I or class II products. Products matching Annex IV are critical products.18

Commission Implementing Regulation (EU) 2025/2392 provides technical descriptions for those categories.19

Core functionality controls classification

Classification concerns the complete product’s principal technical capabilities and intended purpose. Integrating a categorised component does not automatically transfer the component’s classification to the complete product.

A laptop containing a secure element is not automatically a critical product. A news application containing an embedded browser is not automatically class I. The integrated component can remain separately categorised, while its security and integration risks remain relevant to the complete-product assessment.20

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    PRODUCT[Complete product and<br/>its core functionality]:::product
    CATEGORY{Which category applies?}:::test

    DEFAULT[Default category]:::product
    C1[Important class I]:::product
    C2[Important class II]:::product
    CRITICAL[Critical]:::product

    FULL{Relevant harmonised standards<br/>common specifications or eligible<br/>certification fully applied?}:::test
    MANDATORY{Mandatory European<br/>certification adopted?}:::test

    A[Module A<br/>internal control]:::product
    BC[Modules B + C<br/>type examination and<br/>production control]:::product
    H[Module H<br/>full quality assurance]:::product
    CERT[/Applicable European<br/>cybersecurity certification/]:::overlay

    PRODUCT --> CATEGORY
    CATEGORY -- Outside Annex III and IV --> DEFAULT
    CATEGORY -- Annex III class I --> C1
    CATEGORY -- Annex III class II --> C2
    CATEGORY -- Annex IV --> CRITICAL

    DEFAULT --> A
    DEFAULT --> BC
    DEFAULT --> H

    C1 --> FULL
    FULL -- Yes --> A
    FULL -- No --> BC
    FULL -- No --> H

    C2 --> BC
    C2 --> H
    C2 --> CERT

    CRITICAL --> MANDATORY
    MANDATORY -- Yes --> CERT
    MANDATORY -- No --> BC
    MANDATORY -- No --> H

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
Figure 7: Core functionality determines the product category, and the category determines the minimum conformity-assessment route.
Classification Examples in the CRA Minimum available route
Default Most products outside Annexes III and IV Module A; manufacturer may voluntarily choose B+C or H
Important class I Browsers, password managers, anti-malware, VPN products, operating systems, routers, selected security-functional chips and consumer products Module A only under Article 32(2) conditions; otherwise B+C or H
Important class II Hypervisors, container runtimes, firewalls, intrusion systems, tamper-resistant processors and microcontrollers B+C, H, or an available and applicable certification route
Critical Hardware security devices, smart-meter gateways and secure cryptoprocessing devices, smartcards and secure elements Mandatory certification where required by delegated act; otherwise class II routes
Table 2: CRA product classifications and conformity routes.

Module A remains a complete assessment

Under module A, the manufacturer performs the assessment without mandatory notified-body participation. It must still:

  • implement Annex I requirements;
  • conduct adequate examinations and tests;
  • maintain technical documentation;
  • control production and software release;
  • ensure continuing conformity;
  • issue the declaration; and
  • affix CE marking.

Self-assessment is not informal attestation.

Modules B+C and H introduce independent assurance

Under module B, a notified body examines the product type, technical design, development, and vulnerability-handling processes. Module C leaves production conformity with the manufacturer.

Module H instead assesses a full quality system covering design, development, production, testing, and vulnerability handling. It can suit manufacturers with multiple related products or frequent releases, but it is not an unlimited approval for future changes.

Standards create bounded presumptions

A harmonised standard creates a legal presumption only after its reference is published in the Official Journal and only for the requirements it covers.

A technically useful ISO, IEC, ETSI, CEN, or CENELEC document can support the evidence case without having that legal effect. A draft standard, an adopted but uncited European standard, or a supplier claim of alignment does not automatically create Article 27 presumption.

For important class I products, partial application or absence of the relevant harmonised standards, common specifications, or eligible certification can make third-party assessment mandatory. Product-specific risks outside the specification must still be addressed.

CE marking is an output of the chain

CE marking indicates that the manufacturer has completed the applicable conformity procedure and declares compliance with relevant Union harmonisation legislation. It does not mean that the product has no vulnerabilities or cannot be compromised.

The declaration and supporting assessment apply to a defined product state. Changes to software, hardware, remote processing, configuration, or intended purpose can require updated evidence, certificate interaction, or a new conformity decision.

Components, open source, integration, and substantial modification

A final-product manufacturer cannot divide responsibility by pointing to upstream dependencies. Article 13 requires due diligence when integrating third-party components, including non-commercial free and open-source software.21

The due-diligence depth should reflect the component’s:

  • privilege;
  • exposure;
  • security function;
  • access to data and commands;
  • effect on availability or safety;
  • attack surface;
  • substitutability;
  • maintainer capacity; and
  • support horizon.

A component declaration, CE marking, certificate, SBOM, test report, or supplier attestation can reduce uncertainty. It does not prove that the downstream configuration is secure.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    COMPONENT[Third-party component]:::product
    ID[Version, provenance<br/>configuration and build identity]:::product
    RECORD[SBOM or dependency record]:::product
    DUE[Risk-proportionate due diligence]:::product
    CONTROL[Isolation, least privilege<br/>validation and secure configuration]:::product
    TEST[Complete-product verification]:::product
    PRODUCT[[Controlled final-product state]]:::system

    EVIDENCE[/Declaration, certificate<br/>test or supplier evidence/]:::overlay
    OPAQUE[Unsupported or opaque dependency]:::context
    FAILURE{{Unresolved component risk}}:::enforcement

    COMPONENT --> ID
    ID --> RECORD
    RECORD --> DUE
    DUE --> CONTROL
    CONTROL --> TEST
    TEST --> PRODUCT
    EVIDENCE ==> DUE
    OPAQUE -. uncertainty .-> DUE
    DUE -- insufficient --> FAILURE
    TEST -- inadequate treatment --> FAILURE

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef system fill:#f1f3f5,stroke:#5c6770,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
    classDef context fill:#ffffff,stroke:#7a7a7a,stroke-width:1.5px,stroke-dasharray:5 5,color:#404040;
Figure 8: Component evidence supports but does not replace complete-product risk assessment and integration verification.

An SBOM is an inventory, not a verdict

The CRA requires an SBOM in a commonly used, machine-readable format covering at least top-level software dependencies.

An SBOM can identify:

  • package or component names;
  • versions;
  • supplier relationships;
  • hashes or other identifiers; and
  • dependency relationships.

It does not by itself establish:

  • artefact authenticity;
  • runtime reachability;
  • secure configuration;
  • vulnerability absence;
  • exploitability;
  • supportability; or
  • suitability for the downstream product.

Hardware dependencies, build infrastructure, signing services, remote services, and dynamically resolved dependencies can require separate records.

Upstream vulnerabilities remain downstream responsibilities

When a manufacturer identifies a vulnerability in an integrated component, it must report it to the component manufacturer or maintainer and address it in the complete product. Where appropriate, a downstream-developed fix should be shared upstream.

Where no upstream fix exists, the final-product manufacturer may need to:

  • maintain a fork;
  • produce a patch;
  • isolate the component;
  • disable the affected function;
  • add compensating controls;
  • replace the dependency; or
  • redesign the product.

A short upstream support period does not automatically shorten the final product’s declared support obligation.

Open source creates several possible roles

Open-source licensing does not determine CRA status. Responsibility and commercial activity do.

A contributor does not become a manufacturer merely by contributing code to a project outside its responsibility. A non-commercial project can remain outside the ordinary manufacturer regime. A person that commercially places open-source software under its name can be its manufacturer.

An open-source software steward is a legal person that provides sustained support, plays a main role in project viability, and supports software intended for commercial activities. Article 24 gives stewards a tailored regime centred on a documented cybersecurity policy, vulnerability handling, remediation, information sharing, and authority cooperation. A steward does not obtain authority to affix CE marking merely through stewardship.22

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    FOSS[Free and open-source software]:::product
    CONTROL{Commercially placed under<br/>the actor's responsibility?}:::test
    SUSTAIN{Sustained support and<br/>main viability role?}:::test

    FMFR([FOSS manufacturer]):::entity
    STEWARD([Open-source software steward]):::entity
    PROJECT[Contributor or non-commercial project]:::context

    CHANGE[Post-market product change]:::product
    PURPOSE{Changes intended purpose?}:::test
    COMPLIANCE{Affects Annex I Part I<br/>compliance?}:::test
    SUB[Substantial modification]:::product
    MAINTAIN[Maintenance, repair or<br/>non-substantial update]:::product
    REASSESS{{Reassess affected boundary<br/>before market supply}}:::enforcement

    FOSS --> CONTROL
    CONTROL -- Yes --> FMFR
    CONTROL -- No --> SUSTAIN
    SUSTAIN -- Yes --> STEWARD
    SUSTAIN -- No --> PROJECT

    CHANGE --> PURPOSE
    PURPOSE -- Yes --> SUB
    PURPOSE -- No --> COMPLIANCE
    COMPLIANCE -- Yes --> SUB
    COMPLIANCE -- No --> MAINTAIN
    SUB --> REASSESS

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
    classDef context fill:#ffffff,stroke:#7a7a7a,stroke-width:1.5px,stroke-dasharray:5 5,color:#404040;
Figure 9: Open-source responsibility and substantial modification depend on control, commercial supply, and effects on the assessed product state.

Integration and modification begin from different states

Integration before first placement ordinarily creates a new complete product. Substantial modification concerns a product already placed on the market.

A post-market change is substantial where it changes the assessed intended purpose or affects compliance with Annex I, Part I. Release labels such as “patch,” “minor,” or “major” are not decisive.

A small authentication change can affect the complete product. A large visual-interface change can remain immaterial to cybersecurity. The analysis depends on architectural reach, privilege, interfaces, data flows, risk, and intended purpose.

Repairs and spare parts require security-relevant equivalence

Identical spare parts made to the same specifications are excluded under Article 2(6). Compatibility alone is insufficient. A replacement that changes secure boot, cryptography, authentication, protocols, privileges, interfaces, or update behaviour may not be identical for CRA purposes.

The host product also requires a separate modification analysis. A component can be independently regulated while its installation changes the conformity state of the larger product.

What customers, integrators, manufacturers, and importers must check

Verification should begin with a common product record rather than a generic “CRA compliant” field.

That record should identify:

  • product name, model, batch, package, or other identifier;
  • hardware, firmware, and software versions;
  • security-relevant configuration;
  • required remote processing;
  • intended purpose and represented operating environment;
  • manufacturer and importer;
  • conformity route and relevant certificate;
  • EU declaration of conformity;
  • support-period end date;
  • vulnerability contact;
  • update source;
  • integration restrictions; and
  • delivered artefact identity.

Evidence must align with the same product state. A genuine declaration or certificate can be irrelevant where it covers another legal manufacturer, model, hardware revision, configuration, or software version.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TB
    CUSTOMER([Customer]):::entity
    INTEGRATOR([Integrator]):::entity
    MFR([Manufacturer]):::entity
    IMPORTER([Importer]):::entity

    ID[Controlled product identity]:::product
    ROLE[Legal entities and roles]:::product
    EVIDENCE[Risk, conformity and<br/>technical evidence]:::product
    LIFE[Support, update, vulnerability<br/>and decommissioning capability]:::product
    DECISION[Release, import, integration<br/>or acceptance decision]:::product
    FAILURE{{Reject, block or remediate<br/>misaligned assurance}}:::enforcement

    CUSTOMER -. verifies suitability .-> ID
    CUSTOMER -. verifies operability .-> LIFE
    INTEGRATOR -. controls configuration<br/>and product boundary .-> ID
    INTEGRATOR -. assesses change .-> EVIDENCE
    MFR --> ID
    MFR --> ROLE
    MFR --> EVIDENCE
    MFR --> LIFE
    IMPORTER -. verifies before placement .-> ROLE
    IMPORTER -. verifies .-> EVIDENCE

    ID --> ROLE
    ROLE --> EVIDENCE
    EVIDENCE --> LIFE
    LIFE --> DECISION
    ID -- mismatch --> FAILURE
    EVIDENCE -- mismatch --> FAILURE
    LIFE -- unsupported --> FAILURE

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 10: Stakeholder verification connects product identity, legal role, conformity evidence, and lifecycle operability.
Control question Customer Integrator Manufacturer Importer
What product state is involved? Match deployed assets to declaration, instructions, and inventory Record components, versions, configuration, interfaces, and remote dependencies Define the assessed and released artefact Match imported units to declarations and certificates
Which CRA role applies? Identify manufacturer, importer, supplier, and support owner Determine whether work constitutes installation, resale, import, new manufacture, rebranding, or modification Establish responsibility and permissible delegation Confirm its own first-placement role and the third-country manufacturer
What demonstrates conformity? Check declaration, CE marking, support date, and relevant certificate Preserve component evidence and determine whether the complete product needs assessment Maintain risk, testing, technical, classification, and conformity evidence Verify that assessment occurred and documents are retrievable
Can the product be supported? Test updates, recovery, reporting contact, and end-of-support planning Allocate monitoring, testing, approval, deployment, and migration Operate vulnerability handling, update distribution, support, and reporting Maintain a viable manufacturer channel and cessation plan
What reopens the decision? New version, changed environment, end of support, or remote-service change New component, interface, privilege, configuration, branding, or supply model Any change affecting risk, evidence, purpose, classification, or approved type A supplied version or manufacturer state that no longer matches evidence
Table 3: Role-specific CRA verification questions.

Customers should test operability

A customer does not ordinarily repeat the manufacturer’s conformity assessment. It should determine whether the product’s representations and support model fit the deployment.

Useful acceptance checks include:

  • secure commissioning;
  • administrative-access setup;
  • absence or treatment of shared credentials;
  • services enabled by default;
  • installation of an authenticated update;
  • recovery after interrupted updating;
  • reset and secure-data-removal functions;
  • logging and time handling;
  • remote-service failure behaviour; and
  • end-of-support migration.

The CRA does not generally require public disclosure of complete technical documentation or the SBOM. A customer can request additional evidence contractually where its risk justifies it.23

Integrators should control product and role transformations

The integrator should determine whether it is:

  • installing within intended purpose;
  • configuring within supported limits;
  • reselling;
  • importing;
  • creating a new marketed product;
  • supplying manufacturer-controlled remote processing;
  • rebranding; or
  • substantially modifying a product.

The integration baseline should record enabled interfaces, accounts, trust stores, certificates, network rules, remote endpoints, software versions, update authorities, and responsibility for applying security updates.

Manufacturers need a release-readiness gate

Before market placement, the manufacturer should confirm that:

  • the tested artefact is the distributed artefact;
  • the risk assessment covers the release state;
  • Annex I applicability is documented;
  • component due diligence and the SBOM are current;
  • known vulnerabilities have been assessed;
  • secure defaults and updates have been tested;
  • classification and conformity route are correct;
  • certificate conditions remain satisfied;
  • vulnerability and reporting functions are operational;
  • support dates are consistent;
  • Annex II information is complete; and
  • the declaration and CE marking identify the product.

Technical documentation must exist before market placement and be updated where appropriate during the support period.24

Importers must operate a pre-market gate

Before placing a third-country product on the Union market, the importer must verify the manufacturer’s conformity procedure, technical documentation, CE marking, declaration, product identification, instructions, contact information, and support-period disclosure.

The importer should also ensure that required documentation can be produced to authorities. A supplier warranty clause is not a substitute for the statutory verification.

Vulnerability reporting, market surveillance, liability, and transition

Vulnerability handling, mandatory reporting, and market surveillance are separate mechanisms.

A vulnerability can require remediation without meeting the Article 14 reporting trigger. A reportable event can exist before a patch is available. Market-surveillance action can follow non-conformity, significant risk, formal defects, complaints, testing, or information received from cybersecurity authorities.

Article 14 has two reporting triggers

The first trigger is an actively exploited vulnerability: reliable evidence indicates malicious exploitation without the system owner’s permission.

Technical exploitability alone is insufficient. Internal testing, responsible research, or a proof of concept does not by itself establish malicious exploitation.

The second trigger is a severe incident affecting product security. The threshold is met where the incident affects or could affect protection of important data or functions, or leads or could lead to malicious code in the product or a user’s network and information systems.25

Reports are progressive

For an actively exploited vulnerability:

  • early warning: within 24 hours of awareness;
  • vulnerability notification: within 72 hours;
  • final report: within 14 days after a corrective or mitigating measure becomes available.

For a severe incident:

  • early warning: within 24 hours;
  • incident notification: within 72 hours;
  • final report: within one month after the 72-hour notification.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    AWARE[Manufacturer awareness]:::product
    TYPE{Reporting trigger}:::test

    ACTIVE[Actively exploited vulnerability]:::product
    INCIDENT[Severe product-security incident]:::product

    A24[Within 24 hours<br/>early warning]:::product
    A72[Within 72 hours<br/>vulnerability notification]:::product
    AFINAL[Within 14 days after<br/>measure becomes available<br/>final report]:::product

    I24[Within 24 hours<br/>early warning]:::product
    I72[Within 72 hours<br/>incident notification]:::product
    IFINAL[Within one month after<br/>the 72-hour notification<br/>final report]:::product

    USERS([Impacted users]):::entity
    STATUS[/Intermediate status report<br/>where requested/]:::overlay

    AWARE --> TYPE
    TYPE -- Vulnerability --> ACTIVE
    TYPE -- Incident --> INCIDENT
    ACTIVE --> A24
    A24 --> A72
    A72 --> AFINAL
    INCIDENT --> I24
    I24 --> I72
    I72 --> IFINAL
    ACTIVE -. mitigation information .-> USERS
    INCIDENT -. mitigation information .-> USERS
    A72 ==> STATUS
    I72 ==> STATUS

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
Figure 11: Article 14 uses staged reports so that authorities receive an early warning before the investigation is complete.

Awareness should be governed as a time-stamped decision. The manufacturer should define who can establish awareness, who owns the reporting clocks, how product versions and affected Member States are identified, and how incomplete evidence is escalated.

The Single Reporting Platform routes reports

Notifications are submitted through ENISA’s Single Reporting Platform to the coordinating CSIRT associated with the manufacturer’s main Union establishment. Relevant reports are then distributed to coordinating CSIRTs in Member States where the product was made available.

Commission Delegated Regulation (EU) 2026/881 specifies the exceptional cybersecurity grounds under which dissemination can be delayed.26

As of 2 August 2026, Commission and ENISA pages stated that the platform was undergoing testing and was scheduled to become operational by 11 September 2026.2728 Manufacturers should verify the production URL, registration, identity-management, submission formats, and fallback arrangements before Article 14 applies.

Users need actionable information

Authority reporting does not replace user communication. Impacted users—and where appropriate all users—must receive information enabling mitigation.

A useful notice identifies:

  • affected products and versions;
  • exploitation or incident conditions;
  • immediate containment;
  • available updates;
  • recovery precautions;
  • credential, key, or certificate actions;
  • temporary loss of function; and
  • the source of future advisories.

Market surveillance can reach substantive and formal failures

Market-surveillance authorities can request internal documentation needed to assess product design, development, production, vulnerability handling, and conformity. Their analysis is not limited to the CE mark.

Enforcement can address:

  • substantive Annex I or process non-compliance;
  • a nominally compliant product that nevertheless presents significant cybersecurity or public-interest risk; and
  • formal defects such as missing marking, declarations, identification, or technical documentation.

Corrective measures can include bringing the product into conformity, restricting supply, withdrawal, or recall.

Infringement Maximum fixed fine Undertaking-based maximum
Annex I or Articles 13 and 14 €15 million 2.5% of worldwide annual turnover, whichever is higher
Specified economic-operator, conformity, documentation, and cooperation duties €10 million 2% of worldwide annual turnover, whichever is higher
Incorrect, incomplete, or misleading information supplied in response to authority or notified-body requests €5 million 1% of worldwide annual turnover, whichever is higher
Table 4: CRA administrative-fine tiers.

Microenterprises and small enterprises are protected from administrative fines for missing the 24-hour early-warning deadline, but not from the reporting duty or other infringements. Article 64 administrative fines do not apply to open-source software stewards.29

Reporting does not itself increase liability

The mere act of mandatory or voluntary reporting does not subject the notifier to increased liability. That protection does not immunise the underlying defect, late remediation, failure to report, misleading information, or resulting damage.

Directive (EU) 2024/2853 separately modernises product-liability law. It treats software as a product and makes cybersecurity, interconnection, post-market control, recalls, and necessary updates relevant to defectiveness analysis for products placed on the market or put into service after 9 December 2026.30

Transition depends on market events

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    D1[10 Dec 2024<br/>entry into force]:::product
    D2[11 Jun 2026<br/>Chapter IV applies]:::product
    D3[11 Sep 2026<br/>Article 14 applies]:::product
    D4[11 Dec 2027<br/>full application]:::product
    D5[11 Jun 2028<br/>specified transitional<br/>certificates expire]:::product

    LEGACY[Product placed before<br/>11 Dec 2027]:::context
    MOD{Substantially modified<br/>from full application?}:::test
    FULL[Full product requirements<br/>apply to affected boundary]:::product
    REPORT[Article 14 applies<br/>regardless of earlier placement]:::product

    D1 --> D2
    D2 --> D3
    D3 --> D4
    D4 --> D5

    LEGACY --> MOD
    MOD -- Yes --> FULL
    MOD -- No --> REPORT
    D3 --> REPORT

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef context fill:#ffffff,stroke:#7a7a7a,stroke-width:1.5px,stroke-dasharray:5 5,color:#404040;
Figure 12: The CRA transition separates reporting readiness from full product conformity and legacy-product modification.

Products placed on the market before 11 December 2027 are generally subject to the full product requirements only if substantially modified from that date. Article 14 reporting is the major exception: it applies to in-scope earlier products.

A compliant transition programme therefore needs parallel tracks:

  • reporting readiness before September 2026; and
  • product-conformity readiness before December 2027.

A Technical assurance model for procurement and lifecycle governance

The CRA can be implemented as a governed evidence system. The central object is the controlled product state—not the supplier, certificate, contract, or repository in isolation.

Let the product state at time t be:

P_t = \left( H_t, S_t, K_t, D_t, I_t, U_t \right), \tag{1}

where:

  • H_t is the hardware configuration;
  • S_t is the software and firmware configuration;
  • K_t is the security-relevant operational configuration;
  • D_t is the set of external dependencies and remote processing;
  • I_t is the intended purpose and represented environment; and
  • U_t is the support and update state.

Two products with the same commercial name are different assurance objects where one of these elements changes in a security-relevant way.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    STATE[Controlled product state<br/>hardware, software, configuration<br/>dependencies, purpose and support]:::product
    ACTORS([Manufacturer, importer<br/>integrator and customer]):::entity
    ROLES[Role assignments]:::product
    DUTIES[Applicable obligations]:::product
    EVIDENCE[Risk, tests, SBOM<br/>declaration and certificates]:::product
    LIFE[Support, updates, disclosure<br/>reporting and retirement]:::product
    DECISION[Release, import, acceptance<br/>change and retirement decisions]:::product
    EVENT[/Dependency, configuration<br/>threat, support or purpose change/]:::overlay
    FAILURE{{Unsupported assurance claim}}:::enforcement

    ACTORS --> ROLES
    ROLES --> STATE
    STATE --> DUTIES
    DUTIES --> EVIDENCE
    STATE --> LIFE
    EVIDENCE --> DECISION
    LIFE --> DECISION
    EVENT ==> STATE
    EVENT ==> EVIDENCE
    STATE -- identity conflict --> FAILURE
    EVIDENCE -- stale or mismatched --> FAILURE
    LIFE -- unavailable capability --> FAILURE

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 13: A controlled product state links legal roles, obligations, evidence, dependencies, lifecycle capabilities, and decisions.

Assurance is conjunctive

Define the internal assurance predicate:

\mathcal{A}(P_t) = \mathcal{I}(P_t) \land \mathcal{R}(P_t) \land \mathcal{O}(P_t) \land \mathcal{E}(P_t) \land \mathcal{L}(P_t) \land \mathcal{C}(P_t), \tag{2}

where:

  • \mathcal{I} means identity and boundary are controlled;
  • \mathcal{R} means legal roles are correct;
  • \mathcal{O} means obligations are identified;
  • \mathcal{E} means evidence supports the claims;
  • \mathcal{L} means lifecycle capabilities operate; and
  • \mathcal{C} means changes have been assessed.

The conjunction prevents false compensation. More penetration testing does not cure an unidentified legal manufacturer. A detailed declaration does not cure an unusable update channel. A valid certificate does not apply to a different product version.

This model is an internal governance abstraction, not a statutory formula.

Evidence must be typed and bounded

An evidence record should identify:

  • its author or issuer;
  • the product state it covers;
  • covered claims or requirements;
  • version and configuration scope;
  • creation and approval dates;
  • validity or review interval;
  • provenance and integrity;
  • access restrictions;
  • limitations; and
  • superseding evidence.

A declaration answers a different question from an SBOM. A certificate answers a different question from a penetration test. A support statement answers a different question from an update-recovery test.

The organisation should reject evidence from a sufficient set where it is:

  • for the wrong version;
  • stale;
  • unauthenticated;
  • outside the relevant certificate scope;
  • based on incompatible assumptions; or
  • too narrow for the claim.

Procurement should use recurrent gates

Gate Decision Minimum evidence
G0 — Candidate intake Is the candidate sufficiently defined for evaluation? Product description, manufacturer, connectivity, preliminary support
G1 — Requirements Have security, support, integration, reporting, and evidence needs been specified? Intended deployment and measurable requirements
G2 — Selection Is the proposed product state acceptable in principle? Declaration, classification, assessment route, support, update and vulnerability information
G3 — Contract and design freeze Are responsibilities and changes controlled? Product baseline, support date, change notice, dependency and remote-service commitments
G4 — Delivery and acceptance Does the delivered product match and operate as represented? Version verification, secure commissioning, update and recovery tests
G5 — Operation and change Does continued use remain defensible? Advisories, support status, changes, incidents, updates and risk decisions
G6 — Exit Can the product be retired without residual risk? Data removal, credential revocation, dependency removal, evidence archive and replacement plan
Table 5: Procurement and lifecycle assurance gates.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    G0[G0<br/>Candidate]:::product
    G1[G1<br/>Requirements]:::product
    G2[G2<br/>Selection]:::product
    G3[G3<br/>Contract and freeze]:::product
    G4[G4<br/>Delivery and acceptance]:::product
    G5[G5<br/>Operation and change]:::product
    G6[G6<br/>Exit]:::product

    DECISION{Gate satisfied?}:::test
    REMEDIATE{{Reject, remediate<br/>or condition approval}}:::enforcement
    EVENT[/Security-relevant change/]:::overlay

    G0 --> G1
    G1 --> G2
    G2 --> G3
    G3 --> G4
    G4 --> G5
    G5 --> G6
    G2 --> DECISION
    G3 --> DECISION
    G4 --> DECISION
    G5 --> DECISION
    DECISION -- No --> REMEDIATE
    REMEDIATE -. return .-> G1
    EVENT ==> G5
    G5 -- reopen affected gate --> G2

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 14: Procurement assurance recurs from candidate intake through secure retirement.

Contracts should preserve the assurance state

Contracts cannot transfer statutory duties away from the responsible economic operator. They can require cooperation and evidence needed to sustain compliance.

Material provisions can address:

  • exact product and remote-service identity;
  • support end date;
  • update delivery;
  • vulnerability and incident notice;
  • dependency changes;
  • SBOM access where justified;
  • evidence access;
  • manufacturer cessation;
  • remediation and replacement;
  • signing and update continuity;
  • migration and exit assistance; and
  • preservation of regulatory records.

One control plane should connect the organisation

Procurement, engineering, product security, legal, operations, and executive governance should use a shared product-state identifier. The organisation does not necessarily need a new software platform. It does need referential integrity between:

  • asset inventory;
  • supplier records;
  • product versions;
  • role assignments;
  • risk assessments;
  • SBOMs and component records;
  • declarations and certificates;
  • support dates;
  • vulnerability cases;
  • Article 14 reports;
  • change requests;
  • risk acceptances; and
  • retirement records.

The operational test is reconstructability: under time pressure, can the organisation determine what is affected, where it is deployed, who is responsible, which evidence applies, which users must act, and whether continued use remains defensible?

Limits, open questions, and the compliance horizon

The CRA reduces fragmentation but does not eliminate uncertainty.

Five categories remain:

  • legal uncertainty about scope, roles, commercial activity, substantial modification, support, and awareness;
  • technical uncertainty about unknown vulnerabilities, emergent attack paths, and changing dependencies;
  • evidential uncertainty about provenance, version alignment, freshness, and coverage;
  • institutional uncertainty about standards, certification, notified bodies, reporting infrastructure, and enforcement; and
  • operational uncertainty about whether support, updates, reporting, and corrective capabilities will work when required.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TB
    STATE[Controlled product state]:::product
    LEGAL[/Legal uncertainty/]:::overlay
    TECH[/Technical uncertainty/]:::overlay
    EVIDENCE[/Evidential uncertainty/]:::overlay
    INSTITUTION[/Institutional uncertainty/]:::overlay
    OPERATIONS[/Operational uncertainty/]:::overlay

    TEST{Assumptions, limits, owners<br/>and review triggers explicit?}:::test
    DECISION[Conditioned assurance decision]:::product
    BLOCK{{Unsupported reliance}}:::enforcement

    STATE --> LEGAL
    STATE --> TECH
    STATE --> EVIDENCE
    STATE --> INSTITUTION
    STATE --> OPERATIONS
    LEGAL ==> TEST
    TECH ==> TEST
    EVIDENCE ==> TEST
    INSTITUTION ==> TEST
    OPERATIONS ==> TEST
    TEST -- Yes --> DECISION
    TEST -- No --> BLOCK

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef overlay fill:#fff0cf,stroke:#9a6500,stroke-width:1.5px,color:#4d3500;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 15: A defensible CRA decision makes legal, technical, evidential, institutional, and operational uncertainty explicit.

Guidance is useful but non-binding

On 27 July 2026, the Commission published the content of detailed CRA guidance covering scope, open source, substantial modification, support periods, classification, risk assessment, remote processing, reporting, and legislative interaction.31

The accompanying C(2026) 5252 communication states that the annexed content would be formally adopted later when all language versions were available. The guidance also states that it is non-binding and that authoritative interpretation belongs to the Court of Justice.

The FAQs are an even more explicitly provisional Commission-services document. They should be used as interpretive support, not treated as amendments to the Regulation.32

Standards remain in development

The M/606 programme covers horizontal and product-specific CRA standards.33 The Commission’s 2026 rolling plan identified expected dates including:

  • 30 August 2026 for initial horizontal risk-based-development and vulnerability-handling deliverables;
  • 30 October 2026 for product-specific deliverables; and
  • 30 October 2027 for further product-agnostic Annex I work.34

Those dates were still in the future on 2 August 2026. Adoption by a European standardisation organisation is not the last legal step: Official Journal citation is required for Article 27 presumption of conformity.

Manufacturers therefore cannot postpone engineering until all harmonised standards exist. They need risk-based controls and evidence using the Regulation, relevant technical standards, testing, and documented justification.

The conformity ecosystem is developing

The Commission’s conformity-assessment page, updated on 31 July 2026, described CRA notified bodies as appearing in NANDO once available and noted that certification-based routes require further legal specification.35

Manufacturers of class II and critical products should account for:

  • notified-body availability;
  • technical scope;
  • laboratory capacity;
  • lead times;
  • treatment of frequent software changes;
  • certificate revisions; and
  • interaction with other product regimes.

Platform readiness remains a dependency

The Single Reporting Platform was not yet in production on 2 August 2026. Organisations could prepare product inventories, decision roles, EU Login accounts, templates, timestamps, and reporting data, but production operation still had to be verified.

A 24-hour obligation cannot depend on reconstructing product identity, affected territories, or exploit evidence for the first time during an incident.

Conformity does not prove invulnerability

Legal conformity, internal assurance, and actual security are distinct:

  • legal conformity addresses satisfaction of applicable legal duties;
  • internal assurance addresses the evidentiary basis for an organisational decision; and
  • actual security concerns resistance to real threats and configurations.

A risk assessment can omit an attack path. A penetration test is bounded by time and method. An SBOM can omit dynamic behaviour. A certificate can become stale. Absence of reported exploitation can reflect weak detection.

The CRA responds to this epistemic limit through lifecycle processes, not a promise of permanent security.

Dynamic dependencies cause assurance decay

A product can change without a new local executable when:

  • a cloud backend changes;
  • a package resolves to a new artefact;
  • a certificate chain changes;
  • an identity provider modifies behaviour;
  • a component becomes unsupported;
  • an external API changes authentication;
  • a model or dataset changes; or
  • customer configuration exposes a new interface.

Point-in-time evidence must therefore be coupled to monitoring and change assessment.

SMEs face fixed as well as proportional costs

ENISA’s 2026 SME survey received 194 responses from 31 countries or geographical groupings. Sixty-six percent reported prior awareness of the CRA, but practical readiness was weaker. Incident response and lifecycle management were the least mature surveyed domain; more than 70% requested technical-documentation and secure-development templates; and 142 respondents selected financial support as a need.36

The survey was voluntary and is not statistically representative of all EU SMEs. It nevertheless provides evidence that awareness does not automatically produce implementation capacity.

A small manufacturer can still face fixed costs for:

  • secure build and signing systems;
  • independent conformity assessment;
  • vulnerability intake;
  • long-term support;
  • dependency monitoring;
  • reporting readiness; and
  • evidence preservation after staff turnover.

Open questions should be controlled dependencies

Domain Open question as of 2 August 2026 Control
Guidance Has C(2026) 5252 been formally adopted, corrected, translated, or superseded? Preserve the relied-upon version and monitor the Commission publication
Standards Which M/606 deliverables have been adopted and cited? Maintain a standards register with scope and Official Journal status
Notified bodies Which bodies are notified for the product category and module? Verify current NANDO scope before contracting
Certification Has a scheme been made applicable to the relevant CRA route? Verify the enabling act, assurance level, certificate scope, and validity
Reporting Is the production platform operational and what fallback applies? Test access, representatives, formats, timestamps, and contingency channels
Scope Is a remote service, FOSS arrangement, or complex system inside the product boundary? Preserve a case-specific architectural and statutory analysis
Modification Does a change alter intended purpose or affect Annex I compliance? Perform documented impact and substantial-modification review
Support Does the period match expected use and dependency life? Preserve evidence, migration strategy, and end-of-support controls
Awareness When did reliable evidence establish the Article 14 trigger? Time-stamp evidence, uncertainty, decision, and escalation
Overlap Which parallel product, entity, data, safety, or liability laws apply? Maintain a regime-specific obligation and reporting map
Table 6: Controlled CRA implementation questions.
%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    A[2 Aug 2026<br/>verify guidance and readiness]:::product
    B[30 Aug 2026<br/>initial horizontal standards expected]:::product
    C[11 Sep 2026<br/>Article 14 applies<br/>platform scheduled]:::product
    D[30 Oct 2026<br/>product-specific standards expected]:::product
    E[2026–2027<br/>engineering, documentation<br/>and assessment]:::product
    F[30 Oct 2027<br/>additional horizontal<br/>standards expected]:::product
    G[11 Dec 2027<br/>full application]:::product
    H[11 Jun 2028<br/>specified transitional<br/>certificates end]:::product

    A --> B
    B --> C
    C --> D
    D --> E
    E --> F
    F --> G
    G --> H

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
Figure 16: The implementation horizon contains separate decision points for reporting, standards, full conformity, and transitional certificates.

The dates in Figure 16 should be treated as review triggers. Standards, guidance, notified-body listings, certification conditions, and platform procedures can change. Each external dependency should have an owner, monitoring source, latest acceptable decision date, and fallback.

Conclusion: from product features to lifecycle accountability

The CRA changes the unit of cybersecurity governance. The regulated object is not merely a device, software release, supplier, certificate, or feature. It is a product state whose boundary, responsible actors, security basis, dependencies, support commitments, and post-market behaviour must remain traceable.

The resulting compliance chain is shown in Figure 17.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    ID[Controlled product identity]:::product
    ROLES[Correct legal roles]:::product
    REQUIREMENTS[Risk-based security requirements]:::product
    CONTROLS[Implemented and tested controls]:::product
    CONFORMITY[Classification-appropriate conformity]:::product
    PLACEMENT[Secure market placement]:::product
    SUPPORT[Support, remediation<br/>and reporting]:::product
    RETIRE[Controlled modification<br/>or retirement]:::product

    ID --> ROLES
    ROLES --> REQUIREMENTS
    REQUIREMENTS --> CONTROLS
    CONTROLS --> CONFORMITY
    CONFORMITY --> PLACEMENT
    PLACEMENT --> SUPPORT
    SUPPORT --> RETIRE

    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
Figure 17: The CRA compliance chain links controlled product identity to role allocation, risk-based engineering, conformity, market placement, post-market capability, and controlled change or retirement.

Each transition must remain valid. A strong test report for the wrong build is not useful evidence. A certificate does not justify unassessed changes. A component SBOM does not establish complete-product security. A support date does not demonstrate that the update organisation will remain functional.

%%{init: {"theme": "neo", "look": "handDrawn", "layout": "elk"}}%%
flowchart TD
    STATE[Controlled product state]:::product
    ROLES[Legal roles]:::product
    RISK[Risk assessment]:::product
    DESIGN[Secure design, components<br/>testing and updates]:::product
    CONFORMITY[Conformity assessment]:::product
    MARKET[Declaration, CE marking<br/>and market placement]:::product
    SUPPORT[Support and<br/>vulnerability handling]:::product
    RESPONSE[Reporting and<br/>corrective action]:::product
    CHANGE{Security-relevant change?}:::test

    MFR([Manufacturer]):::entity
    IMP([Importer]):::entity
    INT([Integrator]):::entity
    USER([Customer]):::entity
    AUTH([Notified body, CSIRT<br/>or market authority]):::entity

    FAILURE{{Block, restrict, withdraw<br/>recall or retire}}:::enforcement

    STATE --> ROLES
    ROLES --> RISK
    RISK --> DESIGN
    DESIGN --> CONFORMITY
    CONFORMITY --> MARKET
    MARKET --> SUPPORT
    SUPPORT --> RESPONSE
    RESPONSE --> CHANGE
    CHANGE -- Yes --> STATE
    CHANGE -- No --> SUPPORT

    MFR --> STATE
    IMP -. pre-market verification .-> CONFORMITY
    INT -. boundary and configuration .-> DESIGN
    USER -. suitability and operability .-> SUPPORT
    AUTH -. assessment or enforcement .-> RESPONSE

    STATE -- identity unresolved --> FAILURE
    CONFORMITY -- evidence insufficient --> FAILURE
    SUPPORT -- capability unavailable --> FAILURE
    RESPONSE -- correction ineffective --> FAILURE

    classDef entity fill:#dbeaf7,stroke:#285f8f,stroke-width:1.5px,color:#102a43;
    classDef product fill:#dff3e8,stroke:#287a57,stroke-width:2px,color:#153b2b;
    classDef test fill:#ffffff,stroke:#4f5962,stroke-width:1.5px,color:#252a2e;
    classDef enforcement fill:#f8dddd,stroke:#9b2f2f,stroke-width:2px,color:#501717;
Figure 18: CRA compliance is a recurrent lifecycle-accountability loop rather than a permanent status attached to a product name.

For manufacturers, the central capability is a product-assurance system linking release identity, risk, controls, components, tests, conformity, support, updates, and reporting.

For importers, it is a genuine market gate that verifies the foreign manufacturer’s product and evidence before Union placement.

For integrators, it is control of the boundary between installation, new-product manufacture, and substantial modification.

For customers, it is evidence-based procurement and continued-use governance rather than blind reliance on visible markings.

Legal conformity, internal assurance, and actual security remain different propositions. The CRA does not promise permanent invulnerability. It requires manufacturers and supply-chain actors to make cybersecurity risks, decisions, evidence, support, vulnerabilities, and changes governable over time.

The durable compliance capability is therefore reconstructability: the ability to establish what the product is, who is responsible, why it was considered fit for market placement, how it is supported, what changed, who was affected, and what corrective decision followed.

That is the CRA’s fundamental transition—from cybersecurity represented as a collection of product features to cybersecurity governed as lifecycle accountability.

See also cybersecurity longforms

See also regulation and compliance longforms

See also posts

Back to top

Footnotes

  1. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  2. European Parliament and Council of the European Union. (2022). Directive (EU) 2022/2555 on measures for a high common level of cybersecurity across the Union. Official Journal of the European Union. Official text↩︎

  3. European Parliament and Council of the European Union. (2022). Regulation (EU) 2022/2554 on digital operational resilience for the financial sector. Official Journal of the European Union. Official text↩︎

  4. European Parliament and Council of the European Union. (2023). Regulation (EU) 2023/1230 on machinery. Official Journal of the European Union. Official text↩︎

  5. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence. Official Journal of the European Union. Official text↩︎

  6. European Parliament and Council of the European Union. (2024). Directive (EU) 2024/2853 on liability for defective products. Official Journal of the European Union. Official text↩︎

  7. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  8. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  9. European Commission. (2026). Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C(2026) 5252 final and Annex. European Commission. Official publication↩︎

  10. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  11. European Commission. (2026). Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C(2026) 5252 final and Annex. European Commission. Official publication↩︎

  12. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  13. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  14. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  15. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  16. European Commission. (2026). FAQs on the Cyber Resilience Act, version 1.3. European Commission. Official publication↩︎

  17. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  18. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  19. European Commission. (2025). Commission Implementing Regulation (EU) 2025/2392 on the technical description of categories of important and critical products with digital elements. Official Journal of the European Union. Official text↩︎

  20. European Commission. (2026). Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C(2026) 5252 final and Annex. European Commission. Official publication↩︎

  21. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  22. European Commission. (2026). Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C(2026) 5252 final and Annex. European Commission. Official publication↩︎

  23. European Commission. (2026). FAQs on the Cyber Resilience Act, version 1.3. European Commission. Official publication↩︎

  24. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  25. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  26. European Commission. (2026). Commission Delegated Regulation (EU) 2026/881 specifying the terms and conditions for applying cybersecurity-related grounds when delaying dissemination of notifications. Official Journal of the European Union. Official text↩︎

  27. European Commission. (2026). Cyber Resilience Act—Reporting obligations. Shaping Europe’s Digital Future. Official implementation page↩︎

  28. European Union Agency for Cybersecurity. (2026). Single Reporting Platform: Frequently asked questions. ENISA. Official implementation page↩︎

  29. European Parliament and Council of the European Union. (2024). Regulation (EU) 2024/2847 on horizontal cybersecurity requirements for products with digital elements and amending Regulations (EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act). Official Journal of the European Union. Official text↩︎

  30. European Parliament and Council of the European Union. (2024). Directive (EU) 2024/2853 on liability for defective products. Official Journal of the European Union. Official text↩︎

  31. European Commission. (2026). Commission guidance on the application of Regulation (EU) 2024/2847 (Cyber Resilience Act), C(2026) 5252 final and Annex. European Commission. Official publication↩︎

  32. European Commission. (2026). FAQs on the Cyber Resilience Act, version 1.3. European Commission. Official publication↩︎

  33. European Commission. (2026). Cyber Resilience Act—Standardisation. Shaping Europe’s Digital Future. Official implementation page↩︎

  34. European Commission. (2026). Rolling Plan for ICT Standardisation 2026: Cybersecurity and network and information security. Interoperable Europe Portal. Official rolling plan↩︎

  35. European Commission. (2026). Cyber Resilience Act—Conformity assessment. Shaping Europe’s Digital Future. Official implementation page↩︎

  36. European Union Agency for Cybersecurity. (2026). SME CRA Survey Report. ENISA. Official report↩︎

Reuse

Citation

BibTeX citation:
@online{montano2026,
  author = {Montano, Antonio},
  title = {Cyber {Resilience} {Act:} {The} {New} {Product-Security}
    {Baseline}},
  date = {2026-08-03},
  url = {https://antomon.github.io/longforms/cyber-resilience-act-the-new-product-security-baseline/},
  langid = {en},
  abstract = {Regulation (EU) 2024/2847—the Cyber Resilience Act—changes
    European cybersecurity law by applying mandatory, horizontal
    product-security requirements to connected hardware, software, and
    qualifying remote data-processing solutions placed on the Union
    market. Its central innovation is not a particular security control.
    It is the conversion of cybersecurity from a fragmented contractual
    or sectoral concern into an enforceable product-lifecycle
    obligation. This article reconstructs the Regulation from first
    principles. It distinguishes products from services and integrated
    systems; assigns manufacturer, importer, distributor, integrator,
    and user roles; derives the risk-based security baseline in Annex I;
    and explains how product classification determines the permitted
    conformity-assessment route. It examines component due diligence,
    software bills of materials, free and open-source software,
    substantial modification, security updates, support periods, Article
    14 reporting, market surveillance, administrative penalties, and
    interaction with product-liability law. The analysis also develops a
    technical assurance model for procurement and lifecycle governance.
    That model treats the regulated object as a controlled product state
    linked to legal roles, requirements, evidence, dependencies, support
    capability, vulnerability processes, and change decisions. It
    distinguishes legal conformity from internal assurance and from
    actual resistance to attack. The implementation environment remains
    incomplete as of 2 August 2026. Commission guidance is non-binding;
    harmonised standards and notified-body capacity are still
    developing; and ENISA’s Single Reporting Platform is scheduled to
    become operational when Article 14 applies on 11 September 2026.
    These limitations do not suspend the CRA. They make explicit
    assumptions, evidence alignment, monitoring, and controlled
    reassessment essential to defensible compliance.}
}
For attribution, please cite this work as:
Montano, Antonio. 2026. “Cyber Resilience Act: The New Product-Security Baseline.” August 3. https://antomon.github.io/longforms/cyber-resilience-act-the-new-product-security-baseline/.