EU Cyber Resilience Act scoper
Eight questions. Whether the CRA applies to your product, which class it lands in, what you have to produce, and by when.
Assessment
—
—
Timeline
What you have to produce
Gaps against your answers
How the classification works
Scope first
The CRA covers products with digital elements — hardware or software whose intended or reasonably foreseeable use includes a direct or indirect data connection to a device or network. That phrase is doing a lot of work. "Indirect" pulls in devices that never touch a network themselves but exchange data with something that does, and "reasonably foreseeable" means a debug port you did not intend anyone to use still counts.
Out of scope: products already regulated for cybersecurity under sector law — medical devices, type-approved motor vehicles, civil aviation, marine equipment — and products developed exclusively for national security, defense or classified information. Free and open-source software supplied outside a commercial activity is also excluded, though that exclusion is narrower than most maintainers assume once support contracts or paid hosting enter the picture.
Then the class, because it decides who signs off
| Class | Examples | Conformity assessment |
|---|---|---|
| Default | Most connected products | Self-assessment — internal control |
| Important, Class I Annex III | Routers, operating systems, MCUs with security functions, password managers, smart home devices | Self-assessment only if harmonised standards are applied in full — otherwise a notified body |
| Important, Class II Annex III | Firewalls, IDS/IPS, hypervisors, tamper-resistant microcontrollers | Third-party assessment required |
| Critical Annex IV | Hardware security modules, smart meter gateways, smart cards | Third-party, and a European cybersecurity certification scheme may be mandated |
The step that costs money is Class I to Class II. Self-assessment is a documentation exercise you can run internally; a notified body is a external process with lead times measured in months and fees to match. If your product is near that boundary, finding out now is worth considerably more than finding out in 2027.
The essential requirements, in plain terms
Annex I Part I — properties of the product:
- Ship with a secure default configuration, and make reset to that state possible
- No known exploitable vulnerabilities at the point of placing on the market
- Protect confidentiality and integrity of stored, transmitted and processed data — encryption where appropriate
- Process only data that is adequate, relevant and limited to what the function needs
- Protect availability of essential functions, including resilience against denial of service
- Limit attack surfaces, including external interfaces
- Reduce the impact of an incident with mitigation techniques
- Record and monitor security-relevant activity, with opt-out
- Provide security updates — automatic where appropriate, with a clear opt-out
- Allow secure deletion of all data and settings
Annex I Part II — vulnerability handling by the manufacturer:
- Identify and document components, including a machine-readable SBOM of at least the top-level dependencies
- Address and remediate vulnerabilities without delay, including by issuing updates
- Apply effective and regular tests and reviews
- Publicly disclose fixed vulnerabilities once a fix is available, with descriptions and remediation guidance
- Have a coordinated vulnerability disclosure policy, and a contact address for reporting
- Distribute updates securely and without delay, free of charge
Why the update path is the expensive one
Almost every requirement above can be satisfied by documentation and process changes late in a project. The secure update path cannot. It needs flash capacity for two images plus a recovery path, a boot ROM with verified boot, key storage that is not simply a constant in the binary, and monotonic counters to stop a rollback to a vulnerable version. All four are hardware decisions, and all four are fixed by the time the schematic is signed off.
What this tool is not
- Not legal advice. It is a planning aid, and the classification depends on facts about your product that eight questions cannot capture.
- Annex III and IV lists are summarized. Read the actual annexes before making a commercial decision, and take advice on borderline cases.
- Harmonised standards under the CRA are still being developed. The route available to Class I products depends on standards that do not all exist yet.
- Member state enforcement practice will vary, and guidance is still emerging.
Design guidance, not a manufacturing instruction. Verify every result against your fabricator's stack-up and the applicable standard before release. Closed-form models are approximations; the fab's field solver and process window are the authority.
Questions
What people ask about this
When does the Cyber Resilience Act actually bite?
Regulation (EU) 2024/2847 entered into force on 10 December 2024. Three dates matter. Notified-body provisions applied from 11 June 2026. Reporting obligations for actively exploited vulnerabilities and severe incidents apply from 11 September 2026. Everything else — the full set of essential requirements, technical documentation and CE marking — applies from 11 December 2027. A product designed today will be on sale under the full regime, so the architecture decisions being made now are the ones that determine whether compliance is cheap or expensive.
Does the CRA really apply to my little sensor board?
If it has a direct or indirect data connection to another device or a network, and you make it available on the EU market commercially, then yes — almost certainly. The scope is deliberately broad: it is "products with digital elements", not "IoT products" or "software". A wired sensor that talks to a gateway is in scope. So is a firmware image sold separately. The narrow question is not whether you are in scope but which class you land in, because that decides whether you can self-assess.
What is the difference between the CRA and the RED delegated regulation?
They are separate instruments with overlapping intent. The RED delegated regulation (EU) 2022/30 covers radio equipment only, under Article 3(3)(d), (e) and (f) of the Radio Equipment Directive — network protection, privacy, and fraud protection. It has applied since 1 August 2025, so if your product has a radio it is already obligatory today, well before the CRA. The harmonised standards are EN 18031-1, -2 and -3, cited with restrictions that matter in practice. The CRA is broader, applies to everything with digital elements, and arrives in December 2027.
How long do I have to provide security updates?
The support period must reflect the expected product lifetime and be at least five years, unless the product is genuinely expected to be in use for less. For industrial and infrastructure products, five years is a floor and not a target — a meter or a controller installed today may be in service for fifteen. The commercial consequence is what people miss: you are committing to maintain a build environment, a toolchain, a signing infrastructure and the staff to use them for the whole of that period.
What does a secure update mechanism actually require?
At minimum: updates are signed, the device verifies the signature against a key it trusts, the update is delivered over an authenticated channel, and a failed or interrupted update cannot brick the device. In practice that means secure boot, a root of trust that is not a constant in flash, an A/B or recovery partition scheme, and monotonic version counters to prevent rollback to a vulnerable image. This is the single most common gap we find, and it is very expensive to add after the hardware is fixed — the flash budget, the boot ROM options and the key storage all have to be right in the first schematic.
Need this to be right rather than approximately right?
We do CRA gap assessments against an actual design — schematic, firmware architecture, update path and SBOM. Written findings, ranked, with what each one costs to fix now versus later.