News
20 November 2026

CRA regulations: what are your obligations when it comes to managing your vulnerabilities?

Since 11 September 2026, all manufacturers of digital products must report exploited vulnerabilities and serious incidents within 24 hours. What can you do to secure your product? How should you organise the management of CRA vulnerabilities? Rtone can help.

Ensuring the security of your products is part of what we do at Rtone, so we’ve got plenty to tell you about CRA regulations. A little spoiler: the article’s a bit long, and cybersecurity isn’t always the most exciting topic of the week. But it’s for a good cause, and you’ll thank us later 😉

Let’s start with a quick scenario.

Saturday 7 November 2026, 9.00 am. An IT security researcher contacts you: a vulnerability in your product is being exploited. Before 9.00 am on Sunday, an alert must be submitted to the European reporting platform. And it’s up to you to take action.

An unlikely scenario? Not really. It’s one of the requirements of the Cyber Resilience Act (CRA), which has been in force since 11 September 2026. And it doesn’t just apply to your future products: it also applies to those already on the market.

At Rtone, almost all the projects we carry out for our clients fall within the scope of the CRA. We have therefore made it standard practice to integrate these requirements into the project itself – from the build chain right through to the deployed infrastructure – to ensure the product’s security.

In this article, we’ll analyse together whether your product is affected, what your obligations are, how to meet them, and what you need to report. We’ll then detail the six key elements you need to put in place right now to ensure you can comply with confidence.

The CRA: 3 key dates to remember

  • 10 December 2024: entry into force of Regulation (EU) 2024/2847.
  • 11 September 2026: obligation to report actively exploited vulnerabilities and serious incidents.
  • 11 December 2027: full implementation, including cybersecurity requirements and CE marking.

Source: Article 71 of the Regulation, entry into force and implementation

What is the CRA (Cyber Resilience Act)?

Let’s go back to basics. In practical terms, what exactly is the CRA?

It is a European regulation: it applies as it stands in all Member States of the European Union, without the need for national transposition. It sets out the cybersecurity requirements for products sold within the EU and makes CE marking conditional upon compliance.

If the name of the RED Directive rings a bell, you’re not entirely in uncharted territory. The CRA effectively extends its delegated act on cybersecurity: this text, which has been applicable since 1 August 2025, added cybersecurity requirements to the Radio Equipment Directive.

The CRA goes further: it extends these requirements to all digital products, whether they use radio technology or not. And it introduces a major new provision – one that is really worth bearing in mind: the manufacturer becomes responsible for the security of their product throughout its entire lifecycle, and no longer just at the time it is placed on the market. In other words, security no longer ends when the product leaves the factory.

To find out more about the RED, read our article: the RED Directive and cybersecurity

Let’s now move on to THE question you’re no doubt already asking yourself.

 

Is my product covered by the CRA?

The CRA applies to ‘products containing digital elements’ whose intended or reasonably foreseeable use involves a logical or physical connection, whether direct or indirect, to a device or a network (Article 2). This refers to software or hardware products, including their remote data processing solutions and components sold separately (Article 3).

A bit legalistic, we grant you. Let’s break it down 👇

  • Hardware and its embedded software: a product connected to the internet is covered, no question about it. But that’s not all. It is sufficient for any function of the product to require a connection to a network, of any kind, or to another device: updates, retrieving the weather or the time, controlling a computer… Yes, even a USB-connected games controller falls within the scope, as does a port reserved for maintenance.
  • The product’s cloud: this is included if it is designed by the manufacturer, or under their responsibility, and if its absence would prevent the product from performing any of its functions. For example, the cloud that enables a home automation product to be controlled remotely.
  • Software placed on the market: a mobile app, a web browser or a password manager are products in their own right, even without a physical object.

Conversely, a cloud service that does not perform any function of a product, or which is not designed under the manufacturer’s responsibility, does not fall within the scope of the CRA (recital 12). SaaS, PaaS and IaaS offerings, on the other hand, may be subject to the NIS 2 Directive if the organisation providing the service is itself subject to it.

Furthermore, certain products already covered by sector-specific regulations are excluded from the CRA: medical devices and in vitro diagnostic devices, vehicles, civil aviation, marine equipment, identical spare parts, and national security and defence products (Article 2).

💡 The simple rule to bear in mind: if your product does not fall under any of these exclusions and connects to something, it is covered. This is the very purpose of the CRA: to secure as many products as possible.

The scope has been established. Let’s now look at what applies, in practical terms, to your product as of today.

Implementation from 11 September 2026

Some good news to start with: from this date, only the reporting obligations set out in Article 14 are in force. Less good news: they apply to all relevant products, including those placed on the market before December 2027 (Article 69). Your existing deployed fleet is therefore covered.

The text provides for two separate alert mechanisms, each with its own deadlines 👇

Actively exploited vulnerability Severe incident
What does it cover? A flaw in the product for which there is reliable evidence that a malicious actor has exploited it (Art. 3) An event that compromises the availability, authenticity, integrity or confidentiality of important data or functions of the product, or that introduces malicious code (Art. 14)
Early warning Within 24 hours, stating the Member States where the product is sold Within 24 hours, stating whether the incident may have been caused by a malicious act
Notification Within 72 hours: product, nature of the flaw, measures taken or recommended Within 72 hours: nature of the incident, initial assessment, measures taken or recommended
Final report No later than 14 days after a corrective measure is made available Within one month of the notification

 

💡 Key point: a vulnerability that is merely exploitable does not need to be reported. Only a vulnerability that has been exploited, with supporting evidence, is taken into account. However, you are still required to analyse each alert: we’ll show you how to do this in Module 4.

In both cases, you are also obliged to inform the users concerned: you must inform them of the vulnerability or incident, and the corrective or mitigating measures they can take.

What’s the aim of all this? To ensure that every manufacturer does everything possible to secure their product, prevent the exploitation of existing vulnerabilities and ensure that no further vulnerabilities arise.

 

Why is this so important?

You may be wondering whether falling behind on this issue can really have any consequences. The short answer is: yes.

The security of your product is not just a selling point. It is proof of reliability that gives you credibility in your market, at a time when the exploitation of vulnerabilities is on the rise. It is also what makes your customers place their trust in you.

The risks therefore concern your product, of course, but above all your company and its business in the broader sense.

When does the 24-hour deadline start?

Article 14 states that the deadline begins from the moment the manufacturer ‘becomes aware’ of the issue, without defining that moment. Fortunately, the implementation guide published by the European Commission provides further clarification (points 212, 213 and 214) 👇

  • As soon as a report is received (from a customer, researcher, authority or the media), the manufacturer must assess it immediately.
  • The 24-hour period begins when this initial assessment gives the manufacturer reasonable certainty that the vulnerability is being exploited, or that the serious incident has occurred.
  • The duration of this assessment depends on the circumstances, but it must be carried out promptly, especially if the risk is high.

Let’s take a very specific example: a report received on Thursday and left unanswered until Monday puts you in a difficult position. That is why the first step, outlined below, is to appoint someone to receive and assess every alert without delay.

 

What are the risks of non-compliance with the CRA?

Let’s talk about penalties, since we have to. 😅 For a breach of manufacturers’ obligations, the fine can be up to €15 million or 2.5 per cent of annual global turnover – that is, turnover generated across all markets – whichever is higher (Article 64).

And that’s not all: the market surveillance authority may impose other measures, such as the withdrawal or recall of the product in question (Article 54).

There is, however, some flexibility. Micro and small enterprises benefit from a single exemption: they are not subject to financial penalties if they exceed the 24-hour deadline for early warning, whether for a vulnerability or an incident (Article 64, paragraph 10). Please note that all other fines remain applicable, and the reporting obligation remains in full force.

Does this exemption apply to your business? The CRA adopts the European definition (Recommendation 2003/361/EC) 👇

  • Micro-enterprise: fewer than 10 employees, and an annual turnover or balance sheet total of €2 million or less.
  • Small enterprise: fewer than 50 employees, and an annual turnover or balance sheet total of €10 million or less.

Medium-sized enterprises (fewer than 250 employees) do not benefit from this exemption.

The key takeaway: meeting the CRA’s safety requirements protects not only your products but also your business. Rest assured, the aim of this article is not to panic you or lecture you – quite the opposite. Our aim at Rtone is to support you through this process.

So, shall we get started? Let’s look at what needs to be put in place to avoid any risks.

 

The 6 building blocks to put in place to comply with the CRA

Article 14 of the CRA sets out what needs to be done, but not how to do it. It does not, for example, require the creation of a six-person PSIRT (Product Security Incident Response Team). Phew.

So, how should you organise yourselves to report an actively exploited vulnerability or a serious incident within 24 hours, including at weekends? At Rtone, our teams recommend building on six key elements – the very ones we implement in our own projects. Here they are 👇

1. Appoint a lead person and track every alert

It all starts with one person, or a small team, designated by the manufacturer. Their role is similar to that of a DPO under the GDPR.

This person receives internal and external reports, assesses them, decides whether they need to be reported, and drafts the reports. They must be trained in the process and available within 24 hours, including at weekends.

Each alert becomes a ticket in a tracking tool: a GitLab or Jira board is sufficient, with no need for dedicated software or licences. The ticket keeps a record of everything 👇

  • The source: external report, monitoring tool, internal finding.
  • The time of acknowledgement: the time that triggers the 24-hour deadline.
  • Analysis and decision: whether or not to report, and why.
  • Deadlines and reports: 24 hours, 72 hours, final report.

Why such rigour? Because in the event of an audit, this record proves that you have met the deadlines and fulfilled your duty to analyse the issue, even for an alert that was ultimately not reported.

 

2. Publish a Coordinated Vulnerability Disclosure (CVD) policy

A Coordinated Vulnerability Disclosure (CVD) policy allows anyone to report a vulnerability to you without making it public.

The benefit? You gain time to fix the vulnerability before it becomes widely known. And in practice, two elements are all you need 👇

  • A security.txt file: placed in the /.well-known/ folder on your website and any exposed servers, it complies with RFC 9116. It provides a contact address and a PGP key to encrypt communications.
  • A public page, for example /security: this outlines your commitments. Not to take legal action against the researcher, to give them credit, possibly to pay them, and to respond within a specified timeframe (for example, within 24 hours, followed by a weekly update).

The CRA does not set a deadline for publishing a patch. It is standard practice for a researcher to allow the vendor 90 days before disclosing the vulnerability. So make this page easy to find: this is what prevents a vulnerability from being revealed before it is fixed.

Rest assured, you don’t have to start from scratch. ENISA has submitted a technical advice document on security.txt, CVD policies and security advisories for public consultation, complete with templates. Please note that this is still a draft, not an official version. And we support our clients in adapting these templates to their products.

3. Generate the SBOM and monitor vulnerabilities

Are you familiar with the BOM, the bill of materials for your electronic components? The SBOM (Software Bill of Materials) is the same concept for software: it lists all the software components of the product, along with their versions.

It is written in a standard, machine-readable format: CycloneDX or SPDX. ENISA has also published a technical advice document dedicated to the SBOM. It is then compared against public vulnerability databases at two stages:

  • With every build: to ensure that a component that is already vulnerable is not delivered.
  • Every night, throughout the product’s lifecycle: to detect any vulnerabilities discovered after delivery.

How do you actually produce this SBOM? It all depends on the part of the product: the cloud lends itself very well to this, whilst embedded systems require a little more care. We’ll show you how we do it at Rtone.

On the cloud side: an automated pipeline

The cloud relies on package managers, which uniquely identify each dependency. The SBOM is therefore generated almost entirely automatically.

For our Spring Java applications, we have integrated the CycloneDX utility for Maven into the CI/CD pipeline. For our Angular frontends, we use CycloneDX for Node.

Each SBOM is then sent to our instance of Dependency-Track, an open-source software project led by OWASP. It acts as a watchdog: it blocks any release that includes a vulnerable component, then compares the SBOMs every night against the NVD (NIST) and GitHub Advisories databases to alert the relevant manager as soon as a component becomes vulnerable.

Embedded systems: the latest tools, and keeping an eye on archives

With Zephyr, as with Yocto for embedded Linux, the SBOM is generated automatically by the build system.

The difficulty arises with libraries delivered as ZIP archives or copied directly into the code. They do not have a unique identifier, and no tool can detect them on its own. They must therefore be declared manually in the SBOM 👇

  • The name, version and publisher of the library.
  • Its standard identifier where available (PURL or CPE): this is what enables monitoring tools to match it against vulnerability databases.
  • This declaration must be updated with every version change, ideally integrated into the build process to ensure it isn’t overlooked.

And for an older product without these tools? The SBOM is built entirely by hand. Don’t panic – we can take care of it: we’ll inventory the components from the source code and build files, generate the SBOM in the standard format, and then integrate it into daily monitoring.

 

4. Classifying the alert: exploited vulnerability or serious incident?

An alert comes in. Should the ‘red plan’ be triggered? Not necessarily. An alert does not automatically trigger a report: the CRA covers two situations, each following a distinct process with its own timeframes.

For a vulnerability, the person in charge asks three questions 👇

  1. Is the vulnerable component actually in use? A CVE affecting a library function that your product never calls has no impact. You should therefore document that the product is not affected.
  2. Is the vulnerability exploitable in your configuration? A vulnerability that can be exploited remotely as soon as a service is accessible carries less weight than one requiring physical access or the combination of several vulnerabilities.
  3. Do you have evidence of exploitation? Log entries reported by your products, a report accompanied by evidence, or abnormal behaviour observed across your infrastructure. A vulnerability that is trivial to exploit on an exposed service must be considered exploited unless you can prove otherwise.

If the answer to question 3 is ‘yes’, the reporting process begins. Otherwise, the fix follows your usual process, into the backlog, and the decision is recorded in the ticket.

For an incident, the question is whether it is serious within the meaning of the CRA. It often stems from an internal finding: a data leak or tampering, or the unavailability of a service essential to the product. If it affects important product data or functions, or if it introduces malicious code, the incident reporting process begins.

An exploited vulnerability and a serious incident each follow their own process: only the final report differs – 14 days after the patch is applied or one month after notification. Let’s be honest: there is no simple rule that covers every scenario. Hence the importance of a documented analysis, carried out by someone who knows the product.

5. Reporting to ENISA: 24 hours, 72 hours, final report

Reports are submitted via ENISA’s Single Reporting Platform (SRP) and are then forwarded to the coordinating CSIRT in your Member State of establishment. In France, this is ANSSI (CERT-FR).

Good news: the 24-hour early warning is brief. Its purpose is to raise the alert, not to explain everything.

The subsequent reports require more work: measured impact, severity, cause, corrective measures. Hence the importance of laying the groundwork now 👇

  • Create EU Login accounts for authorised personnel, with two-factor authentication. Each account is personal.
  • Read the user guide published by ENISA.
  • Wait to register on the platform: ENISA recommends starting the registration process when the first report is made.

 

6. Fix, deploy and inform users

Reporting is good, but it’s not enough. Fixing it is better. Without a secure over-the-air (OTA) update mechanism, deploying a patch across a device fleet quickly becomes an impossible task.

The deployment time depends on the target and the size of the fleet. For an app, a few minutes may suffice. For an IoT fleet, expect it to take anywhere from a few hours to a few days: updating 10 devices is nothing like updating 10,000.

This is precisely the role of a fleet management platform such as Rtower: to deploy updates remotely and keep track of the status of your fleet.

At the same time, inform the affected users: the vulnerability or incident affecting them, the mitigation measures to be applied even before the patch is installed, and then the release notes detailing what has been fixed.

Through which channel? The CRA does not specify this: email, an in-app notification or a public webpage. For a consumer product, a clear webpage and an in-app notification are often the most suitable options.

For an industrial client managing hundreds of your devices, opt instead for a format that can be read by their tools, such as a safety notice in CSAF format. They will be able to import it directly and identify which devices are affected within minutes.

The importance of simulation before the big day

Once the building blocks are in place, test them. We recommend a dry run: tell the team that a false alarm will be triggered sometime this week, without specifying when. The exercise reveals minor oversights: an EU Login account that hasn’t been activated, a missing contact, an unclear step. And it’s better to discover them on a Tuesday during a drill than on a Saturday during a real crisis.

 

Security by design: what we’ve learnt from real-world experience

At Rtone, we’ve been designing connected devices since 2007. Connecting a product to the internet meant we were confronted with these issues very early on: we’ve made it a habit to design a secure architecture right from the outset, well before the RED or CRA came into force.

We also carry out audits on existing products. Out of around ten gap analyses conducted for the RED and the EN 18031 standard, the same pattern emerges: the gaps are significant, and closing them requires a real effort. Had they been taken into account right from the design stage, they would have required far less effort.

But a concrete example is worth more than a long speech.

The TPM added as an afterthought

A manufacturer of connected devices deployed throughout France commissioned us to audit its product, which had been designed by another team. To meet the requirements, the embedded software’s bootloader needed to be encrypted and its encryption key protected within a TPM, a hardware security chip. There was just one small problem: the board didn’t have one.

Here’s what achieving compliance entailed 👇

  1. Purchase 1,000 TPMs, at a few dollars each.
  2. Fit them onto the 600 circuit boards in stock, with the associated refurbishment costs.
  3. Replace the circuit boards in the 400 products already installed across France: a technician’s call-out and service visit for each one.
  4. Update the embedded software so that it uses the TPM: development, testing and validation.

We’ll leave you to do the maths for your own product… Had it been planned from the design stage, this TPM would have cost a few dollars per board.

This isn’t about pointing the finger at this manufacturer. What happened to them could happen to anyone who hasn’t anticipated cybersecurity requirements, and that is precisely why regulations are being tightened today and why we want to equip you with the tools you need.

December 2027: what’s being added to the CRA

What next? On 11 December 2027, the CRA will apply in full to products placed on the market from that date onwards (Article 71). Two sets of requirements will then be added (Annex I) 👇

  • Security by Design: the cybersecurity requirements that must be met to obtain the CE marking.
  • Vulnerability management: SBOM, coordinated disclosure policy, regular testing, publication of fixed vulnerabilities and free security updates.

The manufacturer must also specify a support period of at least 5 years, unless the product’s intended service life is shorter (Article 13). Technical documentation must be retained for at least 10 years.

Good news: if you’ve followed sections 2, 3 and 6 of this article, you’re already one step ahead.

 

Which category does your product fall into?

Your product’s category determines how you demonstrate its compliance. It depends on its main function, as listed in Annexes III and IV of the Regulation.

The majority of connected devices fall under the default category. However, certain products fall under Class I: smart home virtual assistants, locks, cameras, baby monitors, alarms, connected toys, and health or children’s wearables. For these, self-assessment is only possible by applying a harmonised standard; otherwise, you must go through a notified body.

Note: the harmonised CRA standards, the EN 40000 series, are currently being adopted. The first parts (vocabulary and principles) were approved in 2026; the part on vulnerability management is in the process of being approved; and the part on general requirements is expected by the end of 2027. None have yet been published in the Official Journal.

But there’s no need to panic 👇

  • For the default category, self-assessment remains an option: you can demonstrate directly that your product meets the requirements of Annex I.
  • For Class I, the harmonised standard is a prerequisite for continuing with self-assessment. Without it, you will need to go through a notified body: it is best to factor this into your planning. If the standards are delayed, the Commission may also adopt common specifications, which will serve the same purpose.
  • In the meantime, existing standards already cover a large proportion of the requirements: the RED’s EN 18031 standard, IEC 62443-4-1 and 4-2, or ETSI EN 303 645 for consumer devices. Relying on these today means getting a head start.

At Rtone, we monitor developments in these standards and incorporate them into our projects right from the design stage. We’ll be back very soon with a dedicated article on this topic.

Conclusion

Have you made it this far? Well done – the hard part is over.

The CRA makes cybersecurity a requirement that applies throughout a product’s entire lifecycle. Since 11 September 2026, every manufacturer must be able to detect, assess, report and rectify an exploited vulnerability or a serious incident, including in products already sold.

The six building blocks are within your reach: a designated officer, a CVD policy, a monitored SBOM, a qualification method, a reporting process in place and a secure update process. Integrated from the outset, they have minimal impact on your development process.

And if we were to give you one piece of advice: don’t wait until December 2027. Exploited vulnerabilities are on the rise, and your customers know it. A product with monitored security, and a manufacturer capable of reacting quickly and explaining why, inspires confidence. Beyond compliance, it is not only your product but also your business that you are protecting.

Want to review your products’ CRA compliance? Let’s discuss it

Is your infrastructure already deployed? Find out how Rtower helps you update and monitor it, and discover our cybersecurity expertise.

 

 

 

Special FAQ on the Cyber Resilience Act

Do I need to register on the ENISA platform straight away?

No, ENISA recommends registering when you submit your first report. However, you should create your EU Login accounts with two-factor authentication today, and read the user guide.

Is my product, which will be sold in 2022, affected?

Yes, for the reporting of exploited vulnerabilities and serious incidents, from 11 September 2026. The 2027 requirements only apply in the event of a substantial change, or to units placed on the market after 11 December 2027.

Products already on the market or under development: who must do what, and when?

Falling within the scope of the CRA does not mean that all obligations apply immediately. It all depends on the date the product was placed on the market (Article 69) 👇

  • Products placed on the market before 11 December 2027: these do not need to be brought into compliance with the CRA’s cybersecurity requirements, unless they undergo a substantial modification after that date. However, the reporting obligation has applied to them since 11 September 2026.
  • Products placed on the market from 11 December 2027: the entire Regulation applies (Article 71): security and vulnerability management requirements, technical documentation, conformity assessment, CE marking and reporting.

⚠️ The CRA applies to each individual unit, not to the model. A model that has been on the market since 2025, but of which units are still being manufactured and placed on the market after 11 December 2027, must comply with the requirements for these new units.

In practical terms? 👇

  • Your existing product portfolio: this is already subject to the reporting requirement. You do not need to redesign these products, but you must be able to detect an exploited vulnerability and report it within 24 hours.
  • The robot hoover you have at home: the same principle applies. From September 2026, its manufacturer must report actively exploited vulnerabilities in robots that have already been sold.
  • Your connected bike tracker, currently under development: if it is released after 11 December 2027, it must comply fully with the CRA. You must therefore incorporate these requirements now, at the design stage. If it is released before then, the reporting requirement will apply as soon as it is placed on the market, and units produced after December 2027 must be compliant.
  • An open-source library integrated into your firmware: as a manufacturer, you are liable for the third-party components you integrate (Article 13). If a vulnerability in this library is actively being exploited on your product, it is your responsibility to report it.

Does a public CVE relating to one of my components require me to report it?

Not automatically. The CRA targets vulnerabilities for which there is reliable evidence of exploitation. Analyse the CVE in your specific context and make your decision: if it is not being exploited in your device fleet, it should follow your usual patching process.

Un peu de lecture

Des articles, des podcasts, des webinars… et surtout des conseils pratiques ! En bref, une collection de ressources pour mener à bien votre projet.

Découvrez le cas client : robotique mobile autonome ⭐