The Cyber Resilience Act
The Cyber Resilience Act, in plain English
What the EU's Cyber Resilience Act asks of companies selling products with software in them: who it covers, the dates, what you must produce.
Updated 28 September 2026
On this page
The short version
The Cyber Resilience Act is the EU's new law on products that contain software. Regulation (EU) 2024/2847 of the European Parliament and of the Council on horizontal cybersecurity requirements for products with digital elements (the Cyber Resilience Act).[Title] The Cyber Resilience Act entered into force on 10 December 2024, the twentieth day after its publication in the Official Journal.[Art. 71(1)]
In one paragraph: if you sell a product with software in it in Europe, you have to build it securely, keep a folder of paperwork that shows how, tell your customers how long you will support it, fix security problems while you do, and tell the authorities quickly when a weakness in one of your products is being exploited or something goes seriously wrong with its security. A product that does not meet the rules cannot be sold in the EU.
This page is written for the person who has to organise that work: a managing director, an operations or product manager, an engineering lead. It is our reading of the law as engineers, not legal advice. Where your own position turns on a detail, take advice. Every date and figure on this page comes from our published facts file, each with a link to its source.
Does it apply to your product?
You are very likely covered if all three of these are true.
It has software in it, and it can connect. A product with digital elements is a software or hardware product and its remote data processing solutions, including software or hardware components placed on the market separately.[Art. 3(1)] Machines, building systems, sensors, desktop programs, mobile apps, firmware, developer libraries and server software all count.
The connection is the first question, and it is broader than a network port. The Regulation applies to a product with digital elements whose intended purpose or reasonably foreseeable use includes a direct or indirect data connection — logical or physical — to a device or network. This is the first scoping question, and it is the only test Article 2(1) sets: nothing in it turns on how big the company is or where it is based.[Art. 2(1)] It also counts when the product only reaches a network through something it is part of. A connection counts as indirect when the product does not reach a device or network by itself, but does so as part of a larger system that is directly connectable. That test settles how a product is connected, not who owes what. A component is a product in its own right when it is placed on the market separately; a part a manufacturer builds into its own product and never supplies on its own is covered through the product around it instead. And building in a component that belongs to one of the important or critical categories does not by itself put the surrounding product into that category.[Art. 3(10); Art. 3(1); Art. 7(1)] A sensor board that talks to a gateway, and the gateway talks to the internet, is in.
You sell it, or make money around it, in the EU. Making a product available on the market means supplying it for distribution or use on the Union market in the course of a commercial activity, whether it is paid for or free, and placing on the market is the first such supply. Because the test is supply onto the Union market, a seller established outside the EU is caught in the same way as one inside it, and the size of the company makes no difference. The Regulation does not define "commercial activity" itself; the examples usually quoted for it come from the recitals, which guide interpretation rather than bind.[Art. 3(21), Art. 3(22); recitals for the commercial-activity examples] Selling, licensing and bundling with a paid service all count, wherever your company is based.
A narrow sector carve-out does not already take it out. The sector carve-outs are a short, closed list. Products to which the medical device, the in-vitro diagnostic or the motor vehicle type-approval Regulations apply are outside the Cyber Resilience Act, and so is equipment within the scope of the marine equipment Directive. Civil aviation works differently and more narrowly: what is outside is a product that has actually been certified under the EU civil aviation Regulation, not aviation equipment as a class, so an uncertified product sold into that sector is fully in scope. Outside as well are products developed or modified exclusively for national security or defence purposes — "exclusively" is the whole of that gate, and it asks what the product was developed or modified for rather than who buys it, so neither a product also sold commercially nor a general-market product whose customers today are all defence ones is carved out — and, on a different test, products specifically designed to process classified information. Any wider limitation or exclusion for products covered by other Union rules needs a Commission delegated act; no other rulebook displaces this one by itself. One such act has now been adopted and it takes out a whole sector — the L-category vehicles of Regulation (EU) No 168/2013 — so the closed list above is no longer the whole of the answer, and that sector exclusion has to be read with it.[Art. 2(2), 2(3), 2(4), 2(5), 2(7)] That list is short and closed, and widening it takes a decision by the Commission. Every other EU rulebook — machinery, radio equipment, the AI rules, the network security rules, general product safety — applies alongside this law rather than instead of it. Radio equipment is the clearest example. The Cyber Resilience Act contains no provision that treats meeting its essential requirements as meeting the Radio Equipment Directive's cybersecurity requirements. Recital 30 says only that the Act's requirements include all the elements of the radio-equipment ones, that the Commission may later repeal or amend Delegated Regulation (EU) 2022/30, and that in the meantime it should give guidance to manufacturers caught by both. A recital guides interpretation and does not create a shortcut, so a manufacturer of connected radio equipment has to satisfy both sets of rules today.[Recital 30; recital 100 (a body notified under Delegated Regulation (EU) 2022/30 must be assessed and notified afresh under this Regulation). The Act has 130 recitals; there is no recital 219. There is no enacting article on radio equipment anywhere in the Act. Contrast Art. 12, which is the deeming provision for high-risk AI systems under Regulation (EU) 2024/1689 and has nothing to do with radio equipment] If your product sits on that boundary it needs a review, not an assumption.
Two points catch people out. The first is the cloud half of a product. Remote data processing means processing at a distance for which the software is designed and developed by the manufacturer, or under the manufacturer's responsibility, and without which the product could not perform one of its functions. A pure cloud service that is not part of a product in this sense is not itself covered — that carve-out is stated in the recitals rather than in the definition, and it turns on the responsibility limb: a cloud service designed and developed outside a product manufacturer's responsibility is outside the Regulation.[Art. 3(2); recital 12 for the cloud carve-out] A pure web service with nothing to install is generally outside the law, but the server side of a device that cannot work without it is treated as part of the product.
The second is open source. Free and open-source software developed or supplied outside a commercial activity is not covered. The Regulation does not define "commercial activity"; the examples usually quoted for it — charging for support, monetising the software, a dual licence — come from the recitals, which guide interpretation rather than bind, so whether a given project crosses the line is a question of fact.[Art. 2(1); Art. 3(22); Art. 3(48); recitals for the commercial-activity examples] Whether paid support or a dual licence around an open-source project counts as commercial is a question of fact. Our scope check walks you through both cases in two minutes.
Who has to do what
The law hands out duties by the part you play, not by the size of your company.
The maker. The manufacturer is the person or company that develops or makes a product with digital elements, or has it designed, developed or made by someone else, and sells it under its own name or trademark — whether for payment, for monetisation or free of charge. Outsourcing the build does not move you out of the definition, and neither does giving the product away.[Art. 3(13)] This is the heavy role: build it securely, assess the risk, keep the paperwork, sign the declaration, put the CE mark on it, fix security problems, and report a weakness under attack or a serious incident. If you write the software and sell it under your own name, this is you.
The importer. The EU company that brings in a product made outside the EU. It has to check that the maker did its job before the product is sold.
The distributor. Anyone else in the chain who passes the product on. Lighter duties, but real ones: check that the marks and the documents are there. Importers and distributors check different things, and the distributor's list is the shorter of the two. Before placing a product on the market an importer must make sure the manufacturer carried out the right conformity assessment and drew up the technical documentation, that the product bears the CE marking and comes with the declaration of conformity and the user information, and that the manufacturer has put its identification and contact details on it and stated the end date of the support period. Before making a product available a distributor checks that it bears the CE marking, that the manufacturer and the importer have met their identification, contact-detail, user-information, support-period-end-date and declaration-of-conformity duties, and that all the necessary documents have been handed over. Neither may go ahead where it believes the product is not compliant.[Art. 19(1), 19(2), 19(3); Art. 20(1), 20(2), 20(3)]
Careful: an importer or a distributor can become the maker by accident. An importer or a distributor that places a product on the market under its own name or trademark, or that substantially modifies a product already on the market — a change that affects whether it still meets the product security requirements, or that changes what the product was assessed to be for — is treated as the manufacturer and takes on the manufacturer's product obligations and reporting obligations in full. That deeming rule applies from 11 December 2027 with the rest of the Regulation: Article 71(2) brings forward only Article 14 and Chapter IV, and Article 21 is in neither. A reseller that markets a product under its own name or trademark does not need that rule to be reporting today: the definition of manufacturer covers having a product designed, developed or manufactured and marketing it under one's own name, and Recital 78 says that an entity selling its own branded products qualifies as a manufacturer — so the reporting duty reaches every own-name reseller from 11 September 2026, whether it had the product made for it or buys it in finished. What waits for 11 December 2027 is the substantial-modification limb, for an importer or distributor whose name is not on the product.[Art. 21; Art. 3(13); Art. 3(30); Art. 71(2); Recital 78] Sell a supplier's product under your own brand and the heavy role is yours, reporting included, from the first date rather than the second. Change a product after it is on the market — in a way that affects how secure it is, or that changes what it is for compared with what it was assessed for — and then supply it on, and the maker's duties are yours in full. The narrower version of that rule, the one that stops at the part you changed, is written for someone who is neither the maker, the importer nor the distributor — not for you. Our page for importers and distributors covers this in more detail.
The EU representative. A company outside the EU may appoint an EU-based company to act for it. A manufacturer may appoint an authorised representative in the Union by written mandate; the Regulation does not oblige non-EU manufacturers to appoint one.[Art. 18] We are not one and do not act as one.
The open-source steward. A lighter role for foundations that support open-source projects without selling them. An open-source software steward is a legal person, other than a manufacturer, whose purpose is to provide support systematically and on a sustained basis for the development of specific free and open-source products intended for commercial activities, and that keeps those products viable. Its regime is lighter than a manufacturer's, but it is not empty. A steward must put in place and document a cybersecurity policy, must cooperate with market surveillance authorities on request, and carries part of the reporting duty: notifying an actively exploited vulnerability where the steward is involved in developing the product, and notifying a severe incident and informing users where the incident affects the systems the steward provides for that development. Article 24 is not among the provisions brought forward to 11 September 2026, so on the face of the text those duties start on 11 December 2027 with the rest of the Regulation. No administrative fine may be imposed on a steward for any infringement of the Regulation — but that is true only of the corrected text: as first published, Article 64(10) started at paragraph 3 and left a steward exposed to the top band.[Art. 3(14); Art. 24(1), 24(2), 24(3); Art. 14(1), 14(3), 14(8); Art. 71(2)] Lighter, not empty. A steward keeps a written security policy, works with the authorities, and carries part of the reporting duty itself: for a weakness under attack in a project it helps build, and for a serious incident hitting the systems it provides for that work. Whether that reaches a steward from the first date or only from the second turns on a point the law does not settle cleanly, so if you run a foundation, have it looked at rather than assumed.
Which group is your product in?
Every covered product has to meet the same standard of security. What changes between groups is who checks that you have.
Default products may self-assess their conformity. Important products (Annex III, class I and class II) and critical products (Annex IV) follow stricter routes — with one condition worth knowing. A class I manufacturer that applies in full the harmonised standards, common specifications or a European cybersecurity certification scheme at assurance level at least substantial covering its product is not pushed off the ordinary routes at all, internal control included; it is where those are absent, or applied only in part, that an approved outside body has to be involved. Class II and critical products have no internal-control route on the ordinary path. One exception runs across the whole of Annex III, class II included: a manufacturer of free and open-source software may use the ordinary routes, internal control included, provided the technical documentation is public at the time the product is placed on the market. That exception does not reach the critical products in Annex IV.[Arts. 7, 8, 32; Annexes III, IV]Ordinary products
Anything not on the special lists. Most machines, most business software, most apps. You check your own work, keep the paperwork and sign the declaration yourself. This is where we work, and where most of our customers sit.
Important, class I
Products whose failure has a wider effect. Annex III class I includes, among others, identity and privileged access management software and hardware — including the readers and devices that check a badge, a card, a code or a fingerprint, for access to systems or to buildings — browsers, password managers, anti-malware software, VPN products, network management systems, SIEM systems, boot managers, public-key infrastructure and digital certificate issuance software, network interfaces, operating systems, routers, internet modems and switches, microprocessors, microcontrollers, ASICs and FPGAs with security-related functions, smart home general-purpose virtual assistants, smart home products with security functions such as locks, cameras, baby monitors and alarms, connected toys that come under EU toy safety law and that talk, film or track location, and personal wearables that track health outside medical device law or that are meant to be used by and for children.[Annex III, Class I] You may still check your own work, but only if you apply in full the European standards, or the equivalent rules written for that kind of product. Apply them only in part, or find that none exist, and an outside body has to do the assessment.
Important, class II
A shorter list with a higher security role. Annex III class II covers hypervisors and container runtimes that run operating systems, or similar environments, in virtualised form; firewalls and intrusion detection or prevention systems; and tamper-resistant microprocessors and microcontrollers. The binding technical description of those last two categories is narrower than the heading, so a chip on that boundary has to be read against the description before either conclusion is drawn.[Annex III, Class II] An outside body checks these — with one exception, set out further down this page: an open-source product that publishes its full paperwork can get back to checking its own work.
Critical
A short list where a formal European certificate can be required, but only in the cases the EU has singled out.
Annex IV (critical products) covers hardware devices built around a protected enclosure that resists physical attack, such as payment terminals, hardware security modules and tachographs; smart meter gateways inside smart metering systems, and other devices built for advanced security purposes, including devices that carry out cryptographic processing securely; and smartcards or similar devices, including secure elements such as identity and travel documents, payment and access cards, SIM and eSIM chips, trusted platform modules, qualified signature cards and digital tachograph cards. The binding technical description of the secure-element category is narrower than the heading, so a part on that boundary has to be read against the description before either conclusion is drawn.[Annex IV] A critical product listed in Annex IV needs a European cybersecurity certificate at assurance level at least substantial, and then only where the Commission has adopted a delegated act saying so for that category, and it can only do that where a certification scheme covering the category has already been adopted and is available to manufacturers; any such act must normally allow at least six months to change over, and may allow less only where urgency justifies it. Where no such delegated act has been adopted, the product follows the class II route instead. We have not established which delegated acts exist today.[Art. 8(1); Art. 32(4)]
One thing worth knowing: the European standards that would make this simpler were not finished when we last checked.
There are three routes to a presumption of conformity, not one. A product that conforms to a harmonised standard whose reference has been published in the Official Journal is presumed to conform to the essential requirements that standard covers; so is one that conforms to common specifications laid down in a Commission implementing act; and so is one for which an EU statement of conformity or a certificate has been issued under a European cybersecurity certification scheme, so far as it covers those requirements. Where the Commission has specified a scheme for this purpose, a certificate under it at assurance level at least substantial also removes the need for a third-party conformity assessment for the requirements that certificate covers — not for conformity assessment at large.[Art. 27(1), 27(5), 27(8), 27(9)]As at 10 September 2026, no harmonised standard for the Cyber Resilience Act had had its reference published in the Official Journal, so the presumption of conformity that comes with a harmonised standard was not available and our controls are versioned against the drafts. This is a point-in-time check: it must be re-read and re-dated before it is quoted anywhere, because the day a reference is published the sentence stops being true.[Regulation (EU) 2024/2847 Art. 27(1) — the presumption attaches on publication of the reference in the Official Journal by the Commission, not on adoption by CEN-CENELEC, so the page in source_url cannot evidence an absence from the Official Journal and is the wrong source for this row. The value is the date we last checked, not a figure from the provision.] Until they arrive, your paperwork has to explain your reasoning rather than point at a clause number.
As at 10 September 2026 no harmonised standard had been published for the Cyber Resilience Act, so a class I manufacturer could not reach the internal-control route by that road. That is a point-in-time check and not a permanent state: the day a reference is published, the sentence stops being true. Whether a common specification or a European cybersecurity certification scheme at assurance level at least substantial covers a given category is a separate question we have not established, so check those two before concluding that a notified body is unavoidable.[Art. 32(2)]We work on ordinary products, which is most of what small manufacturers make. If yours turns out to be important or critical, we say so early and tell you what that route involves. We do not run those checks, we are not a certification body, and we will not pretend the extra step away.
The longer version of this section, worked through step by step, is Ordinary, important or critical: which group is your product in?.
The two dates
11 September 2026
The first date is where the reporting duty starts.
Manufacturers' reporting obligations for actively exploited vulnerabilities and severe incidents apply from 11 September 2026.[Art. 71(2); Art. 14] The clocks start when you become aware, not when you finish investigating. Everything you need in order to meet them has to exist before the first incident, not after.
11 December 2027
The second date brings everything else.
The main obligations of the Cyber Resilience Act (essential requirements, technical documentation, conformity assessment, CE marking, importer and distributor duties) apply from 11 December 2027.[Art. 71(2)] From then on: how the product is built, the risk assessment, the paperwork, the declaration, the CE mark, the support period and the security updates. A product that does not meet the rules cannot be put on the EU market.
A simple way to plan. The first date needs a working process for handling security problems. The second needs a documented product. The process is most of the work, and the paperwork falls out of it if you set the process up to leave a trail.
The folder an inspector can ask for
The law calls it the technical documentation. The manufacturer draws up the technical documentation with the content set out in Annex VII before placing the product on the market and keeps it up to date during the support period.[Art. 31; Annex VII] Think of it as the folder you would hand over if someone official said "show me". Nobody stamps it and it is not a certificate. It has to exist before you sell the product and stay current. The manufacturer keeps the technical documentation and the EU declaration of conformity available to market surveillance authorities for at least ten years after the product is placed on the market or for the support period, whichever is longer.[Art. 13(13)]
In ordinary words, it holds:
- What the product is — what it does, which versions, what it runs on, and the information you give the buyer.
- How you build and look after it — the design, how the pieces fit together, how you handle security problems, how updates reach people.
- What could go wrong — the risks you thought about, and how each requirement is met or why it does not apply. Manufacturers carry out and document a cybersecurity risk assessment of the product and include it in the technical documentation.[Art. 13(2), 13(3), 13(4); Annex VII]
- How long you will support it, and why you chose that length.
- Which standards you followed, and what you did where none fitted.
- Test results showing that the product does what you claim and that your process works.
- The signed declaration. The manufacturer draws up and signs the EU declaration of conformity, with the content set out in Annex V, and keeps it up to date; a simplified form is set out in Annex VI.[Art. 28; Annexes V, VI]
- The list of what is inside the product, if it is asked for.
The folder that survives a request is the one where every claim points at something real: a scan report, a change your engineers merged, a signed list of components, a note from a practice run. That is why we build the process first and let it produce the evidence.
The list of what is inside your product
Manufacturers must identify and document the vulnerabilities and the components contained in the product, including by drawing up a software bill of materials in a commonly used and machine-readable format covering at least the top-level dependencies.[Annex I, Part II, point 1]In plain words: a machine-readable list of the software your product is built from, with versions, so tools can read it. The trade calls it a software bill of materials, or SBOM. The law's floor is the components you depend on directly; going deeper is a choice you make because it makes the next duty easier, not because you must. Three things follow.
It has to be produced by your build system, not typed. A hand-written list is wrong within a month and nobody updates it. Have your build produce it on every release, and sign it so a reader can tell it has not been altered.
It is what makes the reporting duty survivable. You cannot learn that a component you ship is being attacked unless you know that you ship it.
It is evidence. Keep each release's list with the paperwork, so you can answer "what was in version 3.2" a year later.
You do not have to publish the list. You have to have it, and produce it if an authority asks. ENISA's June 2026 survey of small and medium firms on the Cyber Resilience Act (194 responses from 31 countries and regional groupings, 174 of them from firms inside the Regulation's scope) found awareness high and readiness low: 67 respondents, about 35%, keep a software bill of materials. ENISA says the survey was voluntary and should not be read as representative of the sector.[ENISA, SME CRA Survey Report, June 2026, executive summary and s.3; DOI 10.2824/0994553; ISBN 978-92-9204-793-1; catalogue TP-01-26-010-EN-N. Every percentage in the report is taken over all 194 responses, not over the 174 in scope] That gap is where most of the readiness work sits.
How to get one, in more detail: Can you say what is inside your product?.
When a weakness is being exploited, or something goes badly wrong
The duty is triggered by knowledge, not by suspicion.
A manufacturer must report an actively exploited vulnerability in its product, and any severe incident having an impact on the security of the product, simultaneously to the CSIRT designated as coordinator and to ENISA.[Art. 14(1), 14(3)] A vulnerability is actively exploited when there is reliable evidence that a malicious actor has exploited it in a system without the system owner's permission. Nothing has to run for that to be true: an authentication bypass, reaching another customer's data, or the use of stolen credentials where a weakness in the product let the attacker obtain or use them, all count, so the 24-hour clock can already be running where no malicious code was involved.[Art. 3(42)]That is narrower than "somebody published a security bug in a library we use". A known weakness that nobody is exploiting goes into your normal queue. A weakness in a product of yours that reliable evidence says a malicious actor has already exploited — in anyone's system, not necessarily one of your customers' — starts the clocks.
The first two filings are due without undue delay and in any event within the outer deadline — the hour figures are backstops, not an allowance to wait. For an actively exploited vulnerability: an early warning without undue delay and in any event within 24 hours of becoming aware, a vulnerability notification without undue delay and in any event within 72 hours, and a final report no later than 14 days after a corrective or mitigating measure is available.[Art. 14(2)]Three practical points:
24 hours
The first message is the hard one. It is short on purpose: what you know, that it is being exploited, which product. It goes out before your analysis is finished.
72 hours
The second message carries substance: what the weakness is in general terms, which product it is in, what you have done and what users should do — and how sensitive you consider the information to be, which is what lets the coordinator hold it back while a fix is still being arranged quietly.
14 days
The last one closes the case once you have a fix: how bad the weakness turned out to be, what you know about whoever exploited it, and the update you shipped.
One narrow piece of relief, if you are a small company: a correction to the law took the fine off the very first deadline for the smallest manufacturers. As the Act was corrected in July 2025, no administrative fine applies to a manufacturer that counts as a microenterprise or a small enterprise for missing the 24-hour deadline on an early warning — either about a weakness under attack or about a serious incident. Because the correction moved the carve-out to start at paragraph 2, it reaches the largest band, 15 million euro or 2.5% of worldwide turnover, which is where a reporting failure sits; before the correction it did not. Read the limits with it: the two 24-hour early warnings only, not a failure to report at all and not the 72-hour notification or the final report; microenterprises and small enterprises only, not medium-sized ones; and it removes the fine, not the duty, so an authority can still order corrective or restrictive measures and a customer or importer will still ask. A second limb of the same paragraph covers open-source software stewards, and it is wider: no administrative fine at all, for any breach of the Act.[Art. 64(10) as corrected by OJ L, 2025/90555, 2.7.2025 ("paragraphs 2 to 9"); Art. 64(10)(a) with Art. 14(2)(a) and Art. 14(4)(a); Art. 64(10)(b) for stewards; Art. 64(2) for the band now reached; Art. 64(9) for fines sitting alongside corrective and restrictive measures] Read it as what it is. The duty to report has not moved, the later deadlines still carry a fine, an authority can still order you to do things, and a customer or an importer will not care that the fine did not apply.
A serious incident affecting the product itself, rather than a weakness in it, is the other half of the duty — and it does not have to be an attack. A bad update that takes out something essential, a misconfiguration that lets one customer see another's data, a failed key rotation: each of them can qualify. It has its own two tests. The first says what kind of incident is covered at all. An incident has an impact on the security of a product when it harms, or is capable of harming, the product's ability to protect the availability, authenticity, integrity or confidentiality of data or functions. This is the gate for what kind of incident the reporting duty is about; whether an incident is severe enough to report is a separate and narrower test.[Art. 3(44)] The second decides whether it is severe enough to report, and it is the test that starts the clock.
An incident that has an impact on the security of the product counts as severe, and so has to be reported, if either of two things is true: it harms, or is capable of harming, the product's ability to protect the availability, authenticity, integrity or confidentiality of sensitive or important data or functions; or it has led, or is capable of leading, to malicious code being introduced or run in the product or in a user's own systems. Either limb on its own is enough, and "capable of" is enough — the duty does not wait for harm that has already happened. This test is for severe incidents only; it never applies to the vulnerability side.[Art. 14(5)]Read that second one carefully before deciding you have nothing to send. It is not the same test as the one for a weakness under attack, and it does not wait for harm that has already happened.
The filings then follow the same shape as for a weakness, with a longer last step. For a severe incident the first two filings carry the same two-part standard: an early warning without undue delay and in any event within 24 hours of becoming aware, and an incident notification without undue delay and in any event within 72 hours. The final report is due within one month after the incident notification is submitted — the month runs from the filing, not from the 72-hour deadline.[Art. 14(4)] The authority that receives your report can also ask for an update in between.
On top of the scheduled filings, the CSIRT that first receives a notification may ask the manufacturer for an intermediate report on where things stand.[Art. 14(6)]The authorities are not your only audience. After becoming aware of an actively exploited vulnerability or a severe incident, the manufacturer must tell the users it affects — and where appropriate all users — about it, and where needed about the steps they can take to limit the damage, in a machine-readable form where that is appropriate. There is no fixed deadline: the standard is that it is done in a timely manner, and if the manufacturer does not, the CSIRT may tell users itself.[Art. 14(8)] Your plan needs a message for customers drafted next to the one for the authorities.
Notifications are submitted through a single reporting platform established by ENISA and run day to day by it, using electronic notification end-points that Member States and ENISA put in place. Two things a manufacturer should know: the reporting duty in Article 14 applies from 11 September 2026, but Article 16, which sets the platform up, is not among the provisions brought forward, so on the face of the text it applies from 11 December 2027; and we have not established whether the platform and the national end-points are operating today. The duty and its deadlines start regardless — where to file in practice is a question for the manufacturer's national CSIRT coordinator.[Art. 16(1); Art. 71(2)] So the practical step this month is to ask your national coordinator how they want to receive a report, and to write the answer down. A notification goes to the CSIRT designated as coordinator in the Member State of the manufacturer's main establishment in the Union, and reaches ENISA at the same time. Main establishment means where the decisions about the cybersecurity of the products are mostly taken — not where the company is registered; if that cannot be worked out, it is where the manufacturer has most of its Union employees. A manufacturer with no main establishment in the Union follows a set order: the Member State of its authorised representative, then of its importer, then of its distributor, then where most of its users are.[Art. 14(7)]
What does not depend on any of that: a named person, a written plan and pre-drafted messages, ready before the first incident. One afternoon spent walking through a pretend incident with the people who would really be in the room is the cheapest way to find out whether the plan works.
What to put in place first, one step at a time: six things to do in a week.
How long you have to look after the product
The manufacturer determines the support period so that it reflects the expected time in use; it must be at least five years unless the product is expected to be used for a shorter time.[Art. 13(8), third subparagraph]You choose the length, you write down why, and you live with it: for the whole period you fix security problems and ship updates. Manufacturers must address and remediate vulnerabilities without delay, including by providing security updates, and where technically feasible those updates must be provided separately from feature updates. Updates must reach users through a secure distribution mechanism and, where that applies to security updates, automatically. Once an update exists it must be sent out without delay, with a message telling users what it covers and what they may need to do, and free of charge — unless the manufacturer and a business customer have agreed otherwise for a product built specifically for that customer.[Annex I, Part II, points 2, 7, 8] Once a fix is out, you also have to say publicly what it fixed. Once a security update has been made available, the manufacturer must share and publish information about the vulnerabilities it fixes: what they are, which products are affected, what the impact and the severity are, and clear instructions for users to remedy them. Publication may be held back, in a case the manufacturer can justify and where the security risk of publishing outweighs the benefit, until users have had the chance to apply the patch. The duty starts when the fix is available, not when the vulnerability is found.[Annex I, Part II, point 4] And you have to tell buyers when support ends. The end date of the support period must be stated clearly to users at the time of purchase and in the information accompanying the product.[Art. 13(19); Annex II, point 7]
One duty outlives the support period, and it is the one people plan around wrongly: every security update you have issued has to stay available long after support has ended. Security updates issued during the support period must remain available for at least ten years after issue or for the remainder of the support period, whichever is longer.[Art. 13(9)] Your paperwork has a long tail of the same kind, so the archive and the download server come down together or not at all. The manufacturer keeps the technical documentation and the EU declaration of conformity available to market surveillance authorities for at least ten years after the product is placed on the market or for the support period, whichever is longer.[Art. 13(13)] Switching the update server off on the last day of support is the mistake to avoid.
Questions worth settling early: how an update reaches a machine already in the field, whether it can be checked and undone, what happens to a customer who never installs it, how long your download and archive have to stay up after support ends, and which parts you depend on will stop being supported before your own period ends.
Does ISO 27001 cover this?
No, but it helps.
ISO 27001 certifies the way your organisation manages security. The Cyber Resilience Act regulates a product: what it does, how it was built and how it is looked after.
If you hold the certificate you already have governance, ownership, incident management and supplier checks, and an auditor who knows your processes. What it does not give you is a risk assessment for the product, a generated list of components, a published way for outsiders to report problems, a chosen support period, a reporting plan keyed to these clocks, or a paperwork folder per product.
If you already hold the certificate the governance work goes faster, and we say so in the quote; it is not a separate discount. Do not expect the certificate on its own to be accepted as evidence for a product.
Selling into the EU from outside
The law follows the product into the EU market, wherever you are. Making a product available on the market means supplying it for distribution or use on the Union market in the course of a commercial activity, whether it is paid for or free, and placing on the market is the first such supply. Because the test is supply onto the Union market, a seller established outside the EU is caught in the same way as one inside it, and the size of the company makes no difference. The Regulation does not define "commercial activity" itself; the examples usually quoted for it come from the recitals, which guide interpretation rather than bind.[Art. 3(21), Art. 3(22); recitals for the commercial-activity examples] A British, American or Asian maker carries the same duties as a French one.
A manufacturer may appoint an authorised representative in the Union by written mandate; the Regulation does not oblige non-EU manufacturers to appoint one.[Art. 18] The mandate says what the representative does; it does not move your responsibility for building a sound product. It does decide something practical, though: if you have no establishment in the EU, the country your reports go to is worked out from the representative first, not from where most of your customers are. A notification goes to the CSIRT designated as coordinator in the Member State of the manufacturer's main establishment in the Union, and reaches ENISA at the same time. Main establishment means where the decisions about the cybersecurity of the products are mostly taken — not where the company is registered; if that cannot be worked out, it is where the manufacturer has most of its Union employees. A manufacturer with no main establishment in the Union follows a set order: the Member State of its authorised representative, then of its importer, then of its distributor, then where most of its users are.[Art. 14(7)]
Expect your EU importers, who have duties of their own, to start asking you for the paperwork, the component list and your reporting arrangements well before the second date.
We are not an EU representative and do not act as one. We do the engineering work in English, and can tell you what a representative does and what that market charges; we have no arrangement with any provider, and choosing one is yours.
Open source in your product
Open source used inside a product you sell is your responsibility like any other component. It goes in the component list, you check what you are pulling in, and you carry the same duties.
Manufacturers exercise due diligence when integrating third-party components, including free and open-source components, so that they do not compromise the security of the product.[Art. 13(5)]The lighter steward role is for organisations that support open-source projects without selling them. An open-source software steward is a legal person, other than a manufacturer, whose purpose is to provide support systematically and on a sustained basis for the development of specific free and open-source products intended for commercial activities, and that keeps those products viable. Its regime is lighter than a manufacturer's, but it is not empty. A steward must put in place and document a cybersecurity policy, must cooperate with market surveillance authorities on request, and carries part of the reporting duty: notifying an actively exploited vulnerability where the steward is involved in developing the product, and notifying a severe incident and informing users where the incident affects the systems the steward provides for that development. Article 24 is not among the provisions brought forward to 11 September 2026, so on the face of the text those duties start on 11 December 2027 with the rest of the Regulation. No administrative fine may be imposed on a steward for any infringement of the Regulation — but that is true only of the corrected text: as first published, Article 64(10) started at paragraph 3 and left a steward exposed to the top band.[Art. 3(14); Art. 24(1), 24(2), 24(3); Art. 14(1), 14(3), 14(8); Art. 71(2)] Lighter is not the same as empty: the reporting duties reach a steward too, in the narrower way set out further up this page.
If you do sell an open-source product and it lands on one of the stricter lists, there is one route back to checking your own work.
A manufacturer of free and open-source software whose product falls in an Annex III category may still use the ordinary routes, internal control included, provided the technical documentation is made public at the time the product is placed on the market. This does not reach the critical products in Annex IV.[Art. 32(5); Art. 31]Being sued, not only fined
The new Product Liability Directive applies to products placed on the market or put into service after 9 December 2026, the date by which Member States must transpose it. Anything placed on the market before that date stays under the 1985 Directive it replaces.[Art. 2(1); Art. 22(1); Art. 21 (Directive 85/374/EEC repealed with effect from the same date but kept for products placed on the market before it)] Directive (EU) 2024/2853, the new Product Liability Directive, counts software as a product. A maker cannot escape a claim by saying the fault only appeared after the product was sold, where the fault comes from the software itself, or from missing software updates needed to keep the product safe, and those were within the maker's control. Cybersecurity requirements count when a court decides whether a product was safe enough. The Directive does not itself oblige anyone to supply updates, and it does not cover free and open-source software developed or supplied outside a commercial activity.[Title; Art. 4(1); Art. 7(1) and 7(2)(f); Art. 11(2)(b) and (c); Art. 2(2); recital 51, last sentence (the Directive imposes no obligation to provide updates or upgrades)]The two laws work together: this one sets what a reasonably secure product looks like, the other is where failing to do it becomes a claim for money. Your paperwork, your decision log and your update history are the evidence in both cases.
Guides to the questions asked most
Each of these takes one question from this page and answers it in full, with a checklist you can work through without us:
- Does the Act cover software and SaaS?
- Default, important or critical: which group is your product in?
- Reporting: what to do in the first week
- The component list: what goes in it, and how to produce it
- How long you must support a product
- Open source: who carries what
- Your name on someone else's product
- Every date that matters
- The fines, and what else can happen
- Importers and distributors: what is yours
Common questions
Does this apply to a small company?
Yes. Duties follow the role you play, not your headcount. Making a product available on the market means supplying it for distribution or use on the Union market in the course of a commercial activity, whether it is paid for or free, and placing on the market is the first such supply. Because the test is supply onto the Union market, a seller established outside the EU is caught in the same way as one inside it, and the size of the company makes no difference. The Regulation does not define "commercial activity" itself; the examples usually quoted for it come from the recitals, which guide interpretation rather than bind.[Art. 3(21), Art. 3(22); recitals for the commercial-activity examples] There is help around the edges, but the duties themselves do not shrink. Member States must, where appropriate, run awareness and training activities about the Regulation, open a dedicated channel for questions, and support testing and conformity assessment — but these measures are aimed at microenterprises and small enterprises, not at medium-sized ones, and each one is qualified by "where appropriate". Conformity assessment fees are a separate matter and do reach medium-sized firms: fees must take account of the interests and needs of micro, small and medium-sized enterprises, including start-ups, and be reduced proportionately.[Art. 33(1); Art. 32(6)] Two of those edges are worth knowing about: a shorter form for the paperwork, and the testing spaces some countries may set up.
A microenterprise or small enterprise may provide the Annex VII technical documentation in a simplified format, and the Commission is required — not merely permitted — to specify that form by implementing act; a firm that takes the simplified route must use the Commission's form, and notified bodies must accept it. Medium-sized enterprises are outside this. It is a simpler form for the same content, not less content, and we have not established whether the form has yet been published.[Art. 33(5); Annex VII]Member States may set up cyber resilience regulatory sandboxes: controlled environments for developing and testing an innovative product against the Regulation for a limited time before it goes on the market. They are optional for Member States, so there may be none where a manufacturer is based. Where one exists, access must be open, fair and transparent, and access for microenterprises and small enterprises must be made easier. A sandbox does not limit what the authorities may require or do.[Art. 33(2)]Our product is software only. Are we covered?
Very likely, if you sell or license it and it can reach a network or a device. Desktop software, mobile apps, developer libraries and server software are all products with software in them.
We only sell a web service. Are we covered?
A pure web service with nothing to install is generally outside this law, though other rules may apply. If your service is the other half of something people install, that half is treated as part of the product.
Remote data processing means processing at a distance for which the software is designed and developed by the manufacturer, or under the manufacturer's responsibility, and without which the product could not perform one of its functions. A pure cloud service that is not part of a product in this sense is not itself covered — that carve-out is stated in the recitals rather than in the definition, and it turns on the responsibility limb: a cloud service designed and developed outside a product manufacturer's responsibility is outside the Regulation.[Art. 3(2); recital 12 for the cloud carve-out]Does a product we already sell have to comply?
Products placed on the market before 11 December 2027 fall under the essential requirements only if they are substantially modified from that date onwards, but the Article 14 reporting obligations are expressly carved out of that relief and apply to every product in scope that was placed on the market before then. From 11 September 2026 the reporting duty therefore covers a manufacturer's whole installed base, not only what it sells from now on.[Art. 69(2), 69(3)] In practice: the reporting duty applies to every product in scope you have ever put on the EU market, whether or not you still support it, not only to what you ship from now on. And a change that affects how secure the product is, or changes what the product is for compared with what was assessed, counts as putting a new product on the market. A substantial modification is a change to a product after it has been placed on the market that either affects whether the product still meets the product security requirements in Annex I Part I, or changes the intended purpose the product was assessed for. A person other than the manufacturer, the importer or the distributor who makes such a change and then makes the product available on the market is treated as the manufacturer. Making available is wider than placing on the market: supplying a product that is already on the market counts, so a systems integrator that reworks someone else's product and passes it on is caught. The obligations that person takes on are the manufacturer's product and reporting obligations for the part of the product the change affects, or for the whole product where the change affects its cybersecurity as a whole. Importers and distributors are reached by their own, separate rule, and that one carries no such limit: an importer or a distributor that places a product on the market under its own name or substantially modifies one already on the market is subject to Articles 13 and 14 for the product, not only for the part it changed. One kind of change is expressly outside the concept, and the statement of it sits in the recitals rather than in the enacting terms: a security update designed to decrease the product's cybersecurity risk, which does not modify the intended purpose, is not a substantial modification, and a minor functionality update such as a visual change or a new interface language generally is not either. A feature update that changes the original intended functions, or the type or performance of the product, is.[Art. 3(30); Art. 22(1), 22(2); Art. 21; Recital 39 for the security-update carve-out, which is intent rather than an operative provision] Get a scoping memorandum per product line rather than guessing.
Do we have to publish the list of what is inside our product?
No. You have to produce it, keep it with your paperwork and hand it over if an authority asks. Some customers will ask for it in a contract; that is a commercial decision, not a legal one.
Who decides whether something has to be reported?
You do, as the maker of the product. Public lists and feeds give you signals; the judgement and the decision to report are yours. A monitoring tool set up in your own account puts the signal in front of your own people, and the runbook we write with you says what they check and what the first message says. The judgement, the decision and the sending stay with you: we are not sent your alerts and we never file anything with an authority.
Can you sign our declaration or certify our product?
No. We are engineers, not a certification body. We do not sign declarations, we do not put the CE mark on anything and we do not act as your EU representative. You remain the maker. We help you produce the evidence.
What should we do first?
Find out what you ship. Get your build system to produce the list of components for each supported product, match it against known weaknesses, name the person who owns reporting, and write down what they would do on day one. Everything else stands on that. The scope check tells you in two minutes whether the law applies to you.
Talk to an engineer
The check says whether this reaches your product, for nothing. Start with the eight steps; what it takes from there, in what order and at what price, is the next question — write to us and we answer within one working day. Or book a call from the last step of the check.