Substantial Modifications and the Support Period: Managing the Compliance Clock Under the CRA

The EU Cyber Resilience Act (CRA) introduces two closely related concepts that, taken together, determine how long a manufacturer is legally on the hook for a product’s cybersecurity: substantial modification and the support period. Treated in isolation, each one feels manageable. Treated together — and especially in iterative software development — they form the single most consequential compliance question for any manufacturer placing software-driven products on the EU market: how long does my obligation to maintain this product actually last, and what triggers a reset of the clock?

This article works through both concepts in light of the European Commission’s March 2026 Draft Guidance on the CRA, with a particular focus on a question we are repeatedly asked: how can a manufacturer prevent a substantial modification from extending the support obligations for older versions of a product still in the field?

The Substantial Modification Test

Article 3(30) of the CRA defines a substantial modification as a change to a product following its placing on the market that either:

  1. affects the product’s compliance with the essential cybersecurity requirements set out in Part I of Annex I, or
  2. results in a modification to the intended purpose for which the product has been assessed.

The two limbs are alternatives, not cumulative. A change can be substantial because it shifts the intended purpose, or because it changes the cybersecurity risk profile in a manner not addressed in the original risk assessment, or both.

Recital 39 of the CRA explains the underlying logic: a change is substantial where it alters the level of cybersecurity risk in a way not considered in the original risk assessment and, as a result, not addressed in the original implementation of the essential requirements.

The Four Practical Criteria

The Commission’s Draft Guidance (paragraph 103) provides the most operationally useful articulation to date. Manufacturers may assess, on a non-exhaustive basis, whether a change:

  • introduces new threat vectors (additional interfaces, communication channels, execution environments, or external dependencies);
  • enables new attack scenarios (new ways unauthorised access, manipulation, or misuse could plausibly occur);
  • changes the likelihood of previously identified attack scenarios (lowering effort or expertise required, increasing exposure, weakening safeguards);
  • changes the potential impact of previously identified attack scenarios (scope of affected data, severity of consequences, ability to detect, contain, or recover).

If none of these apply, and the assumptions and mitigations in the original risk assessment remain valid, the update is unlikely to qualify as substantial. If any of them apply materially, the manufacturer must reassess.

What Is Not a Substantial Modification

Three categories fall clearly outside the substantial-modification threshold:

Security updates. Recital 39 of the CRA and paragraph 101 of the Draft Guidance are explicit: security updates whose primary purpose is to reduce cybersecurity risk are generally not substantial modifications, even where they modify or constrain functionality to address vulnerabilities. Examples 41–43 of the Draft Guidance illustrate this: patching an input-validation bug, tightening firewall rules, disabling unused ports, or rotating a deprecated cryptographic algorithm to a stronger one already supported in the design — none of these reset the clock.

Progressive enabling of foreseen functionality. This is widely misunderstood. If a manufacturer’s original risk assessment explicitly anticipated and assessed a future feature, then enabling that feature later — even years later — is not a substantial modification. Example 38 in the Draft Guidance describes a production-monitoring system shipped with read-only dashboards while automated control features were present but disabled, with the closed-loop control risks already assessed. Enabling those features later did not amount to a substantial modification.

The strategic implication is significant: a thorough, forward-looking original risk assessment is the single most effective way to keep future updates on the safe side of the line.

Physical repairs and identical spare parts. Replacing a defective component with one identical to the original, or with one that performs better but stays within the product’s assessed intended purpose, is not a substantial modification (Draft Guidance paragraphs 89–94).

What Is a Substantial Modification

Conversely, the Draft Guidance makes clear that scale is not the test. Even seemingly minor functional additions qualify if they introduce risks not previously assessed. Example 39 is the canonical case: adding a “remember me” persistent-login feature appears trivial, but it introduces token-theft, unauthorised-access, and session-hijacking risks that were not part of the original risk model. That is a substantial modification.

Likewise, a security update can itself be a substantial modification if it fundamentally alters the trust model. Example 44 in the guidance is instructive: a local file-encryption tool that, in response to a vulnerability, is re-architected to upload files to a remote service operated by the manufacturer. The security objective is genuine, but the product’s intended purpose has materially changed, and a substantial modification has occurred.

Consequences

Where a change is substantial, the modified product is treated as a new product under the CRA. The act of making it available on the EU market constitutes a new placing on the market, and the person carrying out the modification becomes its manufacturer — even if that is the original manufacturer continuing iterative development.

Crucially, this does not mean the entire conformity assessment must be redone from scratch. Paragraph 109 of the Draft Guidance, echoing Section 2.1 of the Blue Guide, confirms that the conformity assessment can focus on the modified parts of the product. Existing documentation and test results for unchanged components remain usable.

For products that were on the market before 11 December 2027, Article 69(2) of the CRA is unforgiving on this point: any substantial modification after that date brings the product fully into CRA scope — provided it would otherwise have benefited from the transitional arrangements. This applies regardless of when the original version was placed on the market.

The Support Period

The support period is the window during which the manufacturer is obliged to handle vulnerabilities effectively, in accordance with the essential requirements of Part II of Annex I. Article 13(8) sets out how the period is determined.

Five Years Is a Floor, Not a Default

A persistent misconception is that “the support period is five years.” It is not. Paragraph 115 of the Draft Guidance puts this beyond doubt:

The minimum support period therefore operates only as a safeguard, ensuring that vulnerabilities are handled for a sufficiently long period, while allowing manufacturers to take into account products with genuinely shorter expected use times. […] A support period of five years is therefore not to be considered as the default for all products.

The reverse is also true: Recital 60 of the CRA explicitly states that products reasonably expected to be in use for longer than five years should have correspondingly longer support periods. The five-year floor only operates downward as a safeguard for users; upward, the period must match the product’s actual expected use time.

Manufacturers determining the appropriate support period are required to consider, among other factors, the product’s intended purpose, the nature of the product, customer expectations, and the lifecycle of similar products in the market. Internal commercial preferences cannot override these criteria.

Communicating the End Date

Two operational obligations follow from Article 13(8):

  • Article 13(19): the end date of the support period (at minimum, month and year) must be indicated clearly and understandably at the time of purchase.
  • Article 13(9): where technically feasible given the nature of the product, the manufacturer must display a notification to users once the support period expires.

These obligations matter both for legal compliance and for the strategic question that follows — managing the support clock when products evolve iteratively.

Where the Two Concepts Collide: The Compliance Clock for Iterative Software

This is the question we are repeatedly asked: a manufacturer places version 1.0 of a software product on the market in 2028 with a declared five-year support period. In 2030, version 2.0 is released and qualifies as a substantial modification. Has the manufacturer just committed to maintaining version 1.0 until 2033 and version 2.0 until 2035 and every intermediate substantially modified version for at least five years each?

If managed naively, yes. If managed carefully, no — and the CRA provides an explicit mechanism for this.

Each Substantially Modified Version Has Its Own Declared Support Period

Paragraph 117 of the Draft Guidance is unambiguous: each version of a software product placed on the market must have a declared support period that complies with Article 13(8) — at least five years, unless the expected use time of that version is demonstrably shorter.

When version 2.0 is placed on the market, the manufacturer must declare a new support period for that version. Version 2.0’s clock starts running from that date. Version 1.0’s originally declared support period does not automatically extend.

Article 13(10): The Lifeline for Iterative Software

This is where most manufacturers either over-commit or under-comply. Article 13(10) creates a specific flexibility for software products:

Where a manufacturer has placed subsequent substantially modified versions of a software product on the market, that manufacturer may ensure compliance with the essential cybersecurity requirement set out in Part II, point (2), of Annex I only for the version that it has last placed on the market, provided that the users of the versions that were previously placed on the market have access to the version last placed on the market free of charge and do not incur additional costs to adjust the hardware and software environment in which they use the original version of that product.

In plain terms: once version 2.0 is on the market and users can upgrade free of charge, you may discontinue active vulnerability remediation for version 1.0 — even before version 1.0’s original support period would have run out.

Paragraph 120 of the Draft Guidance confirms exactly this consequence:

However, manufacturers may rely on Article 13(10) to discontinue addressing and remediating vulnerabilities for earlier versions once users are able to upgrade to a later version free of charge and without incurring additional costs as referred to above, even if this results in a shorter effective support period for those earlier versions.

This is the mechanism. The originally declared support period of version 1.0 is not breached when the manufacturer relies on Article 13(10) to focus vulnerability remediation on version 2.0, provided the conditions are met.

What “Additional Costs” Means

Article 13(10) hinges on users being able to upgrade without “additional costs”. The Draft Guidance (paragraph 119) gives this concept a practical definition:

“Additional costs” does NOT include:

  • reasonable operational effort inherent in applying software updates;
  • personnel time;
  • routine testing;
  • configuration adjustments;
  • upgrades of underlying software dependencies necessary to address end-of-life components or known security vulnerabilities.

“Additional costs” DOES include:

  • mandatory purchases of new hardware;
  • infrastructure replacement;
  • fundamental changes to the operating environment.

In other words: the threshold is met if users can migrate to the new version without procuring new hardware or rebuilding their operating environment. Routine effort to install an update does not qualify as an additional cost.

What Still Applies for Older Versions

Article 13(10) provides relief only from one specific obligation: addressing and remediating vulnerabilities (Annex I Part II point 2). The other vulnerability-handling requirements continue to apply to all subsequent substantially modified versions of the software product placed on the market, as paragraph 120 of the Draft Guidance and Recital 40 of the CRA make clear:

  • maintaining a coordinated vulnerability disclosure policy;
  • maintaining measures to facilitate the sharing of information about potential vulnerabilities;
  • reporting upstream vulnerabilities in integrated components where applicable;
  • where technically feasible, notifying users that their version has reached the end of effective support (Article 13(9)).

Article 13(10) is therefore not a free pass to abandon older versions silently. It is a structured way to concentrate active vulnerability remediation on the current version while keeping the surrounding governance and transparency obligations intact.

The Strategic Playbook: How to Lawfully Shorten the Effective Support Period of Older Versions

For manufacturers releasing substantially modified versions over time, the question is how to operationalise Article 13(10) cleanly and defensibly. The following sequence captures the documented chain of decisions a market surveillance authority will expect to see.

1. Declare the support period for each version at the time it is placed on the market.

Each substantially modified version is a new product. Determine its support period in light of the criteria in Article 13(8) — intended purpose, reasonably foreseeable use, expected use time, and the other factors set out in the regulation. Document the rationale. Communicate the end date in line with Article 13(19).

2. Provide a free upgrade path from the older version.

Article 13(10) requires that users of earlier versions have access to the latest version free of charge. “Free” here means without payment for the software itself; reasonable operational effort to install does not constitute an additional cost (paragraph 119 of the Draft Guidance).

3. Verify that no “additional costs” are imposed in the strict sense.

If the new version genuinely requires new hardware, an infrastructure overhaul, or a fundamental change in operating environment to use, Article 13(10) does not apply. In that case, the manufacturer must continue to address vulnerabilities in the older version for its originally declared support period.

4. Document the decision to discontinue vulnerability remediation for the older version.

The technical documentation should record:

  • when the substantially modified version was placed on the market;
  • that the older version’s users have free access to the new version;
  • that no additional costs (within the meaning of paragraph 119) are imposed;
  • the date from which vulnerability remediation for the older version is being discontinued;
  • the continued application of CVD policy and information-sharing obligations.

5. Notify users.

Article 13(9) requires, where technically feasible, that users of the older version be informed that their version has reached the end of effective support. This is not optional. A version-end notification mechanism in the product, combined with public end-of-support information on the manufacturer’s website, is the standard pattern.

6. Maintain the rest of the vulnerability-handling stack for all versions placed on the market.

CVD policy, the single point of contact under Article 13(17), information-sharing measures, and upstream reporting under Article 13(6) continue to apply across the product line — even for versions where active vulnerability remediation has been discontinued under Article 13(10).

Worked Example

Consider an enterprise software product placed on the market in January 2028 with a declared eight-year support period (running through January 2036). The manufacturer places substantially modified versions on the market in January 2030 and January 2032, each with its own declared eight-year support period. Users of the older versions can upgrade to the latest version free of charge, and applying the upgrade requires only routine configuration adjustments — no new hardware, no infrastructure overhaul.

From January 2030 onward, the manufacturer may, in reliance on Article 13(10), concentrate vulnerability remediation on the 2030 version. The 2028 version no longer receives security patches. From January 2032 onward, vulnerability remediation concentrates on the 2032 version. The 2030 version drops out of active remediation.

Throughout this period:

  • the manufacturer continues to maintain a coordinated vulnerability disclosure policy applicable to all three versions;
  • the manufacturer continues to share information about potential vulnerabilities in all three versions where appropriate;
  • the manufacturer notifies users of the 2028 version that their version has reached end of effective support;
  • the manufacturer keeps the technical documentation up to date to reflect these decisions.

This mirrors Example 47 of the Draft Guidance, which addresses exactly this scenario of iterative version management under Article 13(10). It is fully lawful. The originally declared eight-year support period for the 2028 version is not breached, because Article 13(10) explicitly permits this concentration of remediation effort on the latest version.

What This Does Not Mean

Three guardrails are essential.

It does not mean the support period can be set arbitrarily short for a version that is genuinely expected to be in use for a long time. Article 13(8) and Recital 60 set the floor. A manufacturer cannot declare a one-year support period for a product that is expected to be in field use for ten years.

It does not mean the older version can be left without governance. CVD, information sharing, and user notification obligations continue.

It does not mean every iterative release is a substantial modification. Most security updates and most progressive feature enablements that were foreseen in the original risk assessment do not reset the clock — they are simply maintenance of the version originally placed on the market. The Article 13(10) mechanism applies only when a genuinely substantial modification has occurred and a new version has been placed on the market.

Bringing the Two Concepts Together

The interaction between substantial modification and the support period rewards careful design. A manufacturer that:

  • writes a thorough, forward-looking original risk assessment that anticipates future feature rollouts;
  • maintains clear version control and placement-on-the-market records;
  • communicates support-period end dates accurately at point of purchase;
  • builds an upgrade pipeline that avoids imposing additional costs (in the strict sense) on users;
  • maintains a continuous CVD and information-sharing capability that survives version transitions;

…will be in a position to release substantial updates without inheriting an open-ended set of support obligations across every version ever placed on the market. The clock can be lawfully reset, and the effective support burden on older versions can be lawfully shortened — provided the conditions of Article 13(10) are documented and met.

The manufacturers who will struggle are those who treat each release as either entirely the same product or entirely new, without working through which it actually is. That is precisely the question Article 3(30), Article 13(8), Article 13(10), and Section 4 of the Draft Guidance are designed to answer.

How COMPLYD Helps You Stay Ahead

The compliance questions this article addresses — which changes are substantial, when does a new support clock start, how can older versions be lawfully retired — are precisely the decisions that need to be documented, tracked, and auditable. That is where COMPLYD comes in.

COMPLYD’s platform is purpose-built to simplify CRA compliance and product security management across the lifecycle:

  • Version and Placement Tracking: maintain a clear register of every version placed on the EU market, with declared support periods, intended-purpose statements, and risk-assessment snapshots.
  • Substantial Modification Workflows: walk every product change through the four-criteria assessment from the Commission’s Draft Guidance, with documented rationale for every determination.
  • Support Period Management: model declared and effective support periods across product families, automate end-of-support notifications, and document Article 13(10) reliance.
  • Risk Assessment Workflows: link foreseeable future functionality to the original risk assessment so that progressive feature rollouts do not become unintended substantial modifications.
  • Conformity Support: generate and manage Declarations of Conformity, CE marking documentation, and SBOMs across iterative releases.
  • Expert Guidance: actionable insights and best practices, grounded in the Commission’s Draft Guidance and emerging harmonised standards.

By making the lifecycle clock transparent and manageable, we help you reduce risk, accelerate time-to-market, and ensure your products remain compliant — both at the moment of first placement and across every subsequent iteration.