Skip to content

Cyber Resilience Act SaaS Scope Test: Check Each Component Against the Commission's Guidance

Does the Cyber Resilience Act cover SaaS? Test each component against the Commission's July 2026 guidance and see which reporting duties already apply.

Engineering · 6 Nov 2026 · 11 min read

Cyber Resilience Act SaaS Scope Test: Check Each Component Against the Commission's Guidance

A close-up of a dark blue circuit board with gold contact pads

By@ashtech52Lead Full-Stack Engineer

Does the Cyber Resilience Act apply to your SaaS? A web app reached only through a browser isn't a product on that basis alone, according to the Commission's July 2026 guidance. But a mobile app, desktop client, browser extension or anything else you hand to users can be, and the back-end parts it depends on can then come in with it. Reporting duties for in-scope products have applied since 11 September 2026.

So a Cyber Resilience Act SaaS check runs per component, not per company. Test your installer, agent, SDK, app and back end separately, and you'll probably get different answers for each.

This is general information, not legal advice. The guidance is non-binding, the Commission's cover note calls it a draft awaiting formal adoption, and we haven't checked the Regulation on EUR-Lex directly, so some article wording below comes from an unofficial reproduction and is labelled. Confirm anything you rely on against the official text, and take advice on your own product.

The scope decision is for you and your lawyer. The security work behind it, such as reviewing accounts, cloud and patching, is what our cybersecurity hardening service covers.

Does the Cyber Resilience Act apply to SaaS? Browser-only generally not, installed clients can

The Cyber Resilience Act (CRA) covers "products with digital elements" made available on the EU market. Those are software or hardware products, their remote data processing solutions, and components placed on the market separately.

Here's how that lands for a software vendor:

  • Browser-only SaaS with nothing installed: generally outside, unless it supports the function of a product.
  • SaaS with an installed client: the client can be in scope, and so can the remote processing it relies on.
  • Installed or downloadable software: generally in scope when supplied commercially.

The dates, per the Commission's Communication: in force since 10 December 2024, Article 14 reporting from 11 September 2026, everything from 11 December 2027.

What the Commission's July 2026 guidance says about remote data processing

The Commission published its guidance on 27 July 2026, with 67 examples, use cases and charts. The annex sets out the line in two paragraphs.

Para 20 says software must be supplied to the user and run on the user's side. Para 21 says software that runs remotely and is merely accessed isn't a product on that basis alone, and gives web applications and browser-only progressive web apps as the typical case.

Remote processing then comes in through Article 3(2). Processing at a distance counts as a remote data processing solution (RDPS) when both of these are true:

  • Its absence would stop the product performing one of its functions. The guidance says this isn't limited to core functionality (para 189). Its examples include syncing files, onboarding, configuration, update distribution, and identity and access management.
  • You designed and developed it, or had it built under your responsibility. That means tailor-made for you, not licensed off the shelf (para 195). It doesn't matter who operates it (para 196).

There's also a useful limit. Only the modules responsible for the product's function, and their interfaces, count. Back-end systems the product doesn't directly interact with aren't RDPS, though you still have to assess them as dependencies (paras 205 and 206). Internal systems such as HR, CRM, CI/CD pipelines and penetration testing are out too (para 182).

The result is that your whole estate doesn't come into scope because you shipped an app. A defined slice of it does.

The Cyber Resilience Act SaaS scope test: run these questions on each component

List every component, then ask of each one, in order:

  1. Is it supplied to the user and run on their side? If yes, and it's commercial, treat it as a product.
  2. If not, does one of your products call it directly?
  3. Would that product lose a function without it?
  4. Did you design it, or was it built to your specification?
  5. If you answered yes to 2, 3 and 4, it's RDPS and part of the product. If it's third-party, or the product doesn't reach it directly, assess it as a risk.

Here is how that plays out. "Our reading" marks rows the guidance doesn't state in those words.

Component

Likely result

Basis

Web app used only in a browser, no installed client

Not a product on that basis alone

Para 21, Example 5

Marketing site or help pages

Out, unless they enable a product function

Example 6, para 194

Mobile app from an app store

Product, if commercial

Example 3

Desktop client, including web-technology apps packaged for install

Product, if commercial

Example 4

Browser extension

Product, if commercial

Para 20

Installed agent or CLI

Product, if commercial (our reading)

Para 20 criteria; not named

SDK or library you supply to customers

Likely a separately placed component (our reading)

Paras 17 and 24; see below

API and login service your client calls directly

RDPS, if you designed it or had it built for you

Paras 190, 194 and 195, use case 8.3.1

Ledger, warehouse or job systems behind that API

Not RDPS; assess as a dependency

Paras 205 and 206

Your own app on third-party IaaS or PaaS

May be RDPS

Paras 198 and 199

Third-party SaaS built into your product

Not RDPS; treat like a component if it affects security

Paras 200 and 201

Statistics-only telemetry

Not RDPS; assess the risk

Paras 192 and 193

CI/CD, CRM, HR, pen-test tooling

Not RDPS

Para 182

Public API called only by customers' own code

Not addressed; take advice

No direct text found

SDKs need extra care. An open-source SDK you don't monetise is generally not placed on the market (para 42). But the guidance treats free apps through which a company monetises other services as commercial (para 54, Examples 14 and 15). Our reading is that a free SDK whose purpose is to sell your paid platform could be caught, so don't assume "free" settles it.

A worked example

Take a hypothetical B2B SaaS with a browser app, a mobile app and an on-premises sync agent, all using one API.

  • Browser app: not a product on that basis alone.
  • Mobile app and sync agent: products, each with its own technical documentation entry. Variants with different features can count as distinct products (para 14).
  • API and login service: RDPS for those products, because they call them directly. The same RDPS is declared in each product's file, and the paperwork can be reused (para 204).
  • Reporting pipeline behind the API: not RDPS, because neither client touches it. It goes in the risk assessment.
  • Third-party support chat in the mobile app: not RDPS. Assess it, isolate it and do due diligence, as in the guidance's banking use case.

The practical point: if you drop the mobile app and the agent, the answer for the same back end could change entirely. If that back end is shared between customers, include tenant isolation in the risk assessment; our guide to multi-tenant SaaS on PostgreSQL covers the design side.

Third-party cloud you build on: why IaaS, PaaS and SaaS are treated differently

The test turns on who designed the software, not where it runs. Private clouds and on-premises servers can host RDPS as well (para 187).

Model

What you bring

RDPS?

What you do

IaaS

Your OS and applications

Your software may be; the hypervisor isn't

Document it; check the provider's measures

PaaS

Your application

Your app may be; the provider's OS isn't

Document it; check the provider's measures

SaaS

Configuration only

No

Treat it like a component; assess and do due diligence

"Not RDPS" doesn't mean "ignore". Para 201 expects you to assess the integration risk, mitigate it on your own product and carry out proportionate due diligence. Para 207 lists evidence you can reuse from suppliers, including ISO/IEC 27001:2022 and 27017:2015, and proof of meeting the NIS 2 implementing rules. Para 208 suggests writing security guarantees into your SLAs.

The provider's own major changes don't count as a substantial modification of your product (para 208), but you may need to revisit your risk assessment when they happen.

Cloud services can also carry duties of their own under NIS 2 (para 183), so falling outside the CRA doesn't mean you have no EU cybersecurity duties. DLA Piper makes the same point about NIS 2 and DORA.

Cyber Resilience Act reporting: what has applied since 11 September 2026

If any component is a product, Article 14 applies to it now. This is the part of the CRA that's already live.

The duty covers two events: an actively exploited vulnerability in your product, and a severe incident affecting its security. It extends to your RDPS (para 178). You notify the coordinating CSIRT and ENISA at the same time (para 209), through ENISA's Single Reporting Platform, which went live on 11 September and moved to version 1.2.1 on 20 October 2026.

Stage

Exploited vulnerability

Severe incident

Early warning

Within 24 hours of becoming aware

Within 24 hours of becoming aware

Notification

Within 72 hours of becoming aware

Within 72 hours of becoming aware

Final report

Within 14 days of a fix or mitigation being available

Within one month of the 72-hour notification

Source: guidance para 215. Five details change how you'd prepare:

  • It covers products already on the market. Article 14 applies to in-scope products placed on the market before 11 December 2027, and it continues after support ends (para 210).
  • The clock starts at reasonable certainty. You're aware once an initial assessment gives you a reasonable degree of certainty, and that assessment must happen promptly (paras 213 and 214).
  • There's no retroactive reporting. Exploitation you already knew about before 11 September needn't be reported. Exploitation you learn about after that date must be, even for an old vulnerability (para 217).
  • Third-party flaws can be yours to report. An exploited flaw from a component in your product is reportable by you. One that's unreachable or unexploited in your product isn't, though you should still report it upstream (para 218).
  • You must tell users. Inform impacted users, in proportion to the risk. That doesn't have to mean a public disclosure of details (paras 219 and 220).

An independent guide, cyberresilienceact.eu, notes that the platform doesn't record when you became aware. We haven't checked that on the platform, but keeping your own timestamped record is sensible either way.

The following come from an unofficial reproduction of Article 14, so check them against EUR-Lex:

  • What makes an incident severe (Article 14(5)): it harms, or could harm, the product's ability to keep sensitive or important data or functions available, authentic, intact or confidential; or it has led or could lead to malicious code running in the product or the user's systems.
  • Where non-EU vendors report (Article 14(7)): the Member State of your authorised representative, then importer, then distributor, then wherever you have the most users.

Where the sources still disagree

Core functionality. Check the date on anything that tests only for "core" functions, because many pages ranking for this question predate the guidance. DLA Piper's February 2026 piece frames the test that way, and suggests an online chat feature is likely ancillary. The later guidance says functions aren't limited to core functionality.

In its banking case, a third-party chat is necessary for a function, and escapes RDPS only because someone else built it. An in-house chat might therefore land differently.

How much the guidance counts. It's non-binding, and only the Court of Justice can give an authoritative reading (para 8). There's a further wrinkle. The cover Communication calls the annex draft guidance, to be formally adopted once all language versions exist, and says it applies from then. We couldn't confirm that step has happened.

Installable progressive web apps. Para 21 places browser-only PWAs outside, and para 20 places web-technology apps supplied for local execution inside. A PWA that users install sits between them, and the examples don't settle it. Our reading is that the more it behaves as supplied local software, the more cautious you should be.

Update delivery. Para 182 excludes "distribution of security updates to edge locations" from RDPS, while para 190 lists automated update distribution as a function. Our reading: the function counts, the internal delivery plumbing doesn't.

What to do this quarter, whichever way your test lands

  • Run the component test and write it down. Date it, name who decided, and record why. Case-by-case assessment is what the guidance expects (para 8).
  • Assign an on-call owner for the 24-hour clock. It needs someone who can decide whether an event is reportable, including at weekends.
  • Set up reporting access. Read ENISA's platform guidance and test who can log in before you need to.
  • Write a one-page runbook. Cover the awareness timestamp, the three stages, who drafts what, and how you tell customers.
  • Map your components. You can't tell whether an exploited flaw is in your product without an inventory, and our npm supply chain and LiteLLM write-ups show how quickly that question arrives.
  • Review third parties. Collect the para 207 evidence and add security and change-notice terms to your contracts.
  • If you sell AI features too, check the separate EU AI Act checklist. The Commission says it may issue further guidance on how the two laws interact (para 9).
  • If you're browser-only, record that and set a trigger: any plan for an app, agent, extension or SDK reopens the assessment.

Then plan for 11 December 2027, when the full requirements arrive. Products placed on the market before that date come under the main requirements only if substantially modified afterwards (para 15 and Article 69(2), unofficial copy). New ones need the essential requirements, a conformity assessment and a support period of at least five years, or the expected use time if shorter (para 126). The guidance also covers higher-risk "important and critical" product classes, which we haven't assessed here, so check whether yours is listed.

Frequently asked questions

Is pure SaaS covered by the Cyber Resilience Act?

Generally not on that basis alone. The Commission's guidance says software that runs remotely and is merely accessed isn't a product, which typically covers browser-only web apps. It comes into scope only if it supports the function of a product with digital elements. The guidance is non-binding, and each case needs its own assessment. Other EU rules, such as NIS 2, may still apply.

What is remote data processing under the CRA?

It's processing at a distance that you designed and developed, or had built to your specification, and without which a product couldn't perform one of its functions. It's treated as part of the product. Functions aren't limited to the core ones. Third-party SaaS doesn't qualify, but must be assessed like a component.

Does a mobile app or browser extension for our SaaS count?

Yes, if supplied commercially. The guidance gives app-store mobile apps and browser extensions as software that runs on the user's side. Parts of your back end that the app calls directly can then count as remote data processing for that app, while systems deeper in the stack are assessed as dependencies.

Do reporting duties cover products already on the market?

Yes. Article 14 has applied since 11 September 2026 to all in-scope products, including those placed on the market before 11 December 2027, and continues after support ends. You don't have to report exploitation you already knew about before 11 September 2026.

When do the full requirements apply?

On 11 December 2027. Products placed on the market before then come under the main requirements only if substantially modified afterwards, although reporting already applies to them. New products need to meet the essential requirements, pass conformity assessment and carry a support period of at least five years, unless expected use is shorter.

What are the fines?

Up to €15m or 2.5% of worldwide turnover, whichever is higher, for breaches of the essential requirements and Articles 13 and 14. Lower tiers are €10m or 2%, and €5m or 1%. Micro and small manufacturers aren't fined for missing the 24-hour early warning deadline. These figures come from an unofficial copy of Article 64, so check EUR-Lex.

Got a rough component table or runbook? Book a free consultation, bring the draft, and we'll rank the security work behind it by risk and effort. The legal calls stay with your lawyer.

Sources

Back to all insights

Keep reading.

Have a project in mind?

Tell us what you're building, what's slowing your team down or what you'd like to automate. We'll come back with honest next steps and a clear estimate.

No obligation and no sales script.

Popular searches

Change theme