<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en-GB"><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="https://xyanlabs.com/blog/feed.xml" rel="self" type="application/atom+xml" /><link href="https://xyanlabs.com/" rel="alternate" type="text/html" hreflang="en-GB" /><updated>2026-11-13T13:28:36+00:00</updated><id>https://xyanlabs.com/blog/feed.xml</id><title type="html">Xyan Labs</title><subtitle>We write and maintain the technical files the EU Cyber Resilience Act requires for connected products. We&apos;ve built and shipped connected products ourselves.</subtitle><entry><title type="html">Our product only has a USB port. Is it in scope of the CRA?</title><link href="https://xyanlabs.com/blog/our-product-only-has-a-usb-port/" rel="alternate" type="text/html" title="Our product only has a USB port. Is it in scope of the CRA?" /><published>2026-11-12T00:00:00+00:00</published><updated>2026-11-12T00:00:00+00:00</updated><id>https://xyanlabs.com/blog/our-product-only-has-a-usb-port</id><content type="html" xml:base="https://xyanlabs.com/blog/our-product-only-has-a-usb-port/"><![CDATA[<p>Almost certainly. The CRA applies to products whose intended purpose or reasonably foreseeable use includes “a direct or indirect logical or physical data connection to a device or network” (<a href="https://eur-lex.europa.eu/eli/reg/2024/2847/oj/eng"><code class="language-plaintext highlighter-rouge">Art. 2(1)</code></a>). A USB cable to a laptop is a physical data connection to a device. The test doesn’t need a network, a radio or the internet.</p>

<h2 id="what-the-text-says">What the text says</h2>

<p>A physical connection is one made “through electrical, optical or mechanical interfaces, wires or radio waves” (<code class="language-plaintext highlighter-rouge">Art. 3(9)</code>). A USB port is an electrical interface on the end of a wire. Recital 9 lists “files” among the ways a product can be connected logically. So a memory card that carries firmware or data counts as well.</p>

<p>The Commission’s guidance of July 2026 doesn’t mention USB, memory cards or offline products. It does say where the line sits. Scope is “anchored not in the mere presence of electronics, but in the capacity of a product with digital elements to exchange digital information” (paragraph 26). A signal used only “to trigger or power a function, without conveying digitally encoded information” isn’t a data connection (paragraph 28).</p>

<h2 id="when-a-usb-port-doesnt-bring-a-product-in">When a USB port doesn’t bring a product in</h2>

<p>A port wired for power only exchanges no digital information. A USB-C socket that charges a torch, with its data lines connected to nothing that reads them, doesn’t bring the torch into scope on its own.</p>

<p>Check the schematic, not the connector. If the data lines reach the microcontroller, we’d treat the port as a data connection, even if today’s firmware ignores them. The test is reasonably foreseeable use, and a firmware update can change what the port does. If you conclude a port is power-only, write down why.</p>

<h2 id="an-example-from-our-own-work">An example from our own work</h2>

<p>We recently looked at a handheld camera with no radio and no network stack. Its only links to anything else are a USB-C port and a memory card. Over USB, the firmware picks what the camera appears as: a storage drive, a webcam or a printer link (PictBridge). Updates are a file copied onto the card.</p>

<p>Scope was an easy call: both are data connections. The update file was the interesting part. We took the published firmware images apart, and each one ends in a two-byte checksum: the sum of every byte before it. There’s no signature. Anyone can change the image, correct those two bytes, and the camera installs it.</p>

<p>That matters more than it first looks, because of the USB port. The shipped firmware has no keyboard mode. But the chip’s USB stack picks its device class in software. A modified image, built with the chip vendor’s development kit, could make the camera present itself to a laptop as a keyboard and type commands.</p>

<p>The requirement this touches is integrity: protect programs and configuration “against any manipulation or modification not authorised by the user” (<code class="language-plaintext highlighter-rouge">Annex I Part I (2)(f)</code>). The fix is to sign updates and have the bootloader check the signature. That closes the keyboard route as well.</p>

<p>The rest of the camera’s risk assessment is short. With no network, many of the essential requirements have little to act on. The technical file has to say why for each one that doesn’t apply (<code class="language-plaintext highlighter-rouge">Art. 13(4)</code>), which takes a paragraph each, not a chapter.</p>

<h2 id="if-your-product-looks-like-this">If your product looks like this</h2>

<ul>
  <li><strong>Write the scope reasoning down.</strong> Which ports carry data, what they connect to, and why any power-only port is power-only.</li>
  <li><strong>Sign firmware updates,</strong> however they arrive: over USB, on a card or in the factory.</li>
  <li><strong>Ask your chip vendor for the SDK’s components and versions.</strong> The camera’s firmware had no version strings in it, so the SBOM had to come from the factory, not the binary.</li>
  <li><strong>Check the class.</strong> Most USB-only products are default class and can be self-assessed. The exception is a product whose core function is on the lists in <code class="language-plaintext highlighter-rouge">Annex III</code> or <code class="language-plaintext highlighter-rouge">Annex IV</code>; a hardware security key may be one.</li>
  <li><strong>Reporting already applies.</strong> A product like this on sale today is covered by the <a href="/blog/article-14-reporting-is-live/">Article 14 reporting duty</a> now.</li>
</ul>

<p>If you aren’t sure which side of the line your product sits, the <a href="/cra-checker/">CRA checker</a> asks the same questions. Our <a href="/services/#assessment">CRA assessment</a> puts the answer in writing.</p>]]></content><author><name></name></author><summary type="html"><![CDATA[Almost certainly. A USB port that carries data is a physical data connection, and the CRA covers those whether or not the product ever reaches a network. A port that only charges is a different case.]]></summary></entry><entry><title type="html">Article 14 reporting is live: what the 24-hour clock asks of you</title><link href="https://xyanlabs.com/blog/article-14-reporting-is-live/" rel="alternate" type="text/html" title="Article 14 reporting is live: what the 24-hour clock asks of you" /><published>2026-10-24T00:20:00+01:00</published><updated>2026-10-24T07:50:58+01:00</updated><id>https://xyanlabs.com/blog/article-14-reporting-is-live</id><content type="html" xml:base="https://xyanlabs.com/blog/article-14-reporting-is-live/"><![CDATA[<p>The first CRA duties are live. Since 11 September 2026, a manufacturer that becomes aware of an actively exploited vulnerability in its product, or a severe incident that affects the product’s security, has to notify it within 24 hours (<code class="language-plaintext highlighter-rouge">Art. 14</code>).</p>

<p>This applies to every product with digital elements in scope, including products placed on the market before the rest of the regulation applies on 11 December 2027 (<code class="language-plaintext highlighter-rouge">Art. 69(3)</code>). If you have connected products on sale in the EU today, the clock already applies to them.</p>

<p>The Commission’s <a href="https://digital-strategy.ec.europa.eu/en/library/commission-publishes-new-guidance-support-timely-cyber-resilience-act-implementation">guidance of July 2026</a> adds that it carries on after a product’s support period has ended (paragraph 210). It isn’t retroactive: exploitation you knew about before 11 September needn’t be reported, but anything you learn of now must be, however old the product (paragraph 217).</p>

<h2 id="the-three-deadlines">The three deadlines</h2>

<p>Reports go through the single reporting platform to the CSIRT designated as coordinator in the member state of your main establishment, and to ENISA.</p>

<table>
  <thead>
    <tr>
      <th>Report</th>
      <th>Exploited vulnerability</th>
      <th>Severe incident</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Early warning</td>
      <td>24 hours after you become aware</td>
      <td>24 hours after you become aware</td>
    </tr>
    <tr>
      <td>Notification</td>
      <td>72 hours after you become aware</td>
      <td>72 hours after you become aware</td>
    </tr>
    <tr>
      <td>Final report</td>
      <td>14 days after a fix or mitigation is available</td>
      <td>One month after the notification</td>
    </tr>
  </tbody>
</table>

<p>Every deadline runs from the moment you become aware. That makes the time you learned something a piece of evidence in itself. A log that records when a report arrived, who triaged it and when it was judged exploited is what lets you show you met the clock.</p>

<p>You also have to tell the users affected, and where needed tell them how to mitigate it (<code class="language-plaintext highlighter-rouge">Art. 14(8)</code>). The Commission’s guidance reads this as risk-based: it doesn’t have to mean a public announcement (paragraphs 219 to 221).</p>

<h2 id="when-the-clock-starts">When the clock starts</h2>

<p>The guidance puts the moment you become aware at the point where an initial assessment gives you “a reasonable degree of certainty” that the vulnerability is being exploited (paragraph 213). A vulnerability in a component that isn’t exploitable in your product isn’t reportable (paragraph 218).</p>

<p>So the record should show both: when a report arrived, and when your assessment reached that degree of certainty. It also shows why a report you didn’t make wasn’t due.</p>

<h2 id="using-the-reporting-platform">Using the reporting platform</h2>

<p>ENISA’s <a href="https://www.enisa.europa.eu/news/the-cra-single-reporting-platform-is-launched">single reporting platform</a> went live on 11 September 2026. A few practical points from its FAQ:</p>

<ul>
  <li><strong>You need an EU Login account with two-factor authentication</strong> to register as the person who submits. ENISA suggests registering only when you need to submit. We’d set up accounts for whoever would submit now, so the first hour of a 24-hour clock isn’t spent on a login.</li>
  <li><strong>There’s no API yet.</strong> Reports are entered in a web form, so tooling can prepare the content but not send it.</li>
  <li><strong>If the platform is down,</strong> ENISA says to submit once it’s back. You can contact your CSIRT directly in the meantime, but the notification still has to go through the platform.</li>
</ul>

<h2 id="what-to-have-in-place-before-the-first-one">What to have in place before the first one</h2>

<ul>
  <li><strong>A way to become aware.</strong> Published contact details, such as an email address, for vulnerability reports to reach you, automated monitoring of the components in your SBOM, and a route from support and field teams into security.</li>
  <li><strong>A decision record.</strong> “Actively exploited” is a judgement. Write down who makes it, on what evidence, and record each call with its time.</li>
  <li><strong>Named people.</strong> Who drafts, who approves and who submits, with cover for weekends and holidays. The 24 hours are calendar time, not working hours.</li>
  <li><strong>Prepared content.</strong> Product identification, the member states where it is available, and templates for each stage, so the early warning process is easy to complete within the deadline, ideally taking just minutes.</li>
  <li><strong>A rehearsal.</strong> Run one past incident through the process end to end and time it.</li>
</ul>

<h2 id="where-this-fits-in-the-file">Where this fits in the file</h2>

<p>We appreciate many manufacturers will not yet have suitable product/process documentation in place, but this obligation under the CRA is a great time to get started.</p>

<p>The reporting process belongs in the vulnerability handling part of your technical documentation (<code class="language-plaintext highlighter-rouge">Annex VII (2)</code>), and every report you make belongs in the product’s record. A notification that cannot be traced back to the vulnerability log and the release that fixed it is hard to defend later.</p>

<p>Reach out for help with creating your process.</p>]]></content><author><name></name></author><summary type="html"><![CDATA[Since 11 September 2026, manufacturers must report actively exploited vulnerabilities and severe incidents within 24 hours, for products already on the market too.]]></summary></entry><entry><title type="html">Is your product in scope of the CRA? Five questions that decide it</title><link href="https://xyanlabs.com/blog/is-your-product-in-scope/" rel="alternate" type="text/html" title="Is your product in scope of the CRA? Five questions that decide it" /><published>2026-10-24T00:10:00+01:00</published><updated>2026-10-24T07:50:58+01:00</updated><id>https://xyanlabs.com/blog/is-your-product-in-scope</id><content type="html" xml:base="https://xyanlabs.com/blog/is-your-product-in-scope/"><![CDATA[<p>Most teams ask “are we in scope?” as one question. The regulation answers it as five, and the order matters. You can work through the same questions in our <a href="/cra-checker/">CRA checker</a>.</p>

<h2 id="1-is-it-a-product-with-digital-elements">1. Is it a product with digital elements?</h2>

<p>Software, or hardware with software or firmware in it, including its remote data processing solutions (<code class="language-plaintext highlighter-rouge">Art. 3(1)</code>). A purely mechanical product is out. Almost anything with a microcontroller is in.</p>

<h2 id="2-does-it-connect-directly-or-indirectly">2. Does it connect, directly or indirectly?</h2>

<p>The CRA applies where the product’s intended purpose or reasonably foreseeable use includes a direct or indirect data connection to a device or network (<code class="language-plaintext highlighter-rouge">Art. 2(1)</code>). “Indirect” does a lot of work here. A device that is only ever updated over USB, or that talks to a phone app that talks to the internet, is connected.</p>

<h2 id="3-is-it-made-available-on-the-eu-market-in-a-commercial-activity">3. Is it made available on the EU market in a commercial activity?</h2>

<!-- vale Google.GenderBias = NO -->
<!-- "Open-source software steward" is the CRA's own term (Art. 3(14)). -->
<p>Selling into the EU counts, online sales included. Free products count too when they are monetised. Open-source software published outside a commercial activity is outside the regulation for its developers. An organisation that systematically supports open-source software for commercial use may be an open-source steward, with lighter duties (<code class="language-plaintext highlighter-rouge">Art. 24</code>).
<!-- vale Google.GenderBias = YES --></p>

<h2 id="4-does-a-sector-regulation-take-it-out">4. Does a sector regulation take it out?</h2>

<p>Medical and in vitro diagnostic devices, type-approved vehicles, certified civil aviation products and marine equipment have their own cybersecurity rules and are excluded (<code class="language-plaintext highlighter-rouge">Art. 2</code>). So are products developed exclusively for national security or defence.</p>

<p>The exclusion covers the regulated product only. An app, gateway or accessory sold separately from an excluded device can still be in scope on its own.</p>

<h2 id="5-what-is-the-core-function">5. What is the core function?</h2>

<p>If you are in scope, class decides the conformity route. A product is important (<code class="language-plaintext highlighter-rouge">Annex III</code>) or critical (<code class="language-plaintext highlighter-rouge">Annex IV</code>) only if its <strong>core functionality</strong> is one of the listed categories (<code class="language-plaintext highlighter-rouge">Art. 7(1)</code>). A smart thermostat that runs an embedded operating system is not an operating system. A router is a router.</p>

<table>
  <thead>
    <tr>
      <th>Class</th>
      <th>Typical route</th>
    </tr>
  </thead>
  <tbody>
    <tr>
      <td>Default</td>
      <td>Self-assessment, Module A</td>
    </tr>
    <tr>
      <td>Important, Class I</td>
      <td>Module A with harmonised standards applied in full; otherwise a notified body</td>
    </tr>
    <tr>
      <td>Important, Class II</td>
      <td>A notified body, Modules B and C or Module H</td>
    </tr>
    <tr>
      <td>Critical</td>
      <td>A European cybersecurity certificate where required</td>
    </tr>
  </tbody>
</table>

<h2 id="where-answers-need-writing-down">Where answers need writing down</h2>

<p>The quick cases are quick. The ones that take an argument are products with several functions near an Annex III category, products whose cloud service may or may not be remote data processing, and products that pass through several hands. For those, the reasoning matters as much as the answer: it’s the first thing a market surveillance authority asks for.</p>

<p>Writing that reasoning down is the first part of our <a href="/services/#assessment">CRA assessment</a>: scope and class in writing, with the argument for any borderline call.</p>]]></content><author><name></name></author><summary type="html"><![CDATA[Scope under the Cyber Resilience Act turns on five questions. Most products answer them quickly; the ones that do not need a written rationale.]]></summary></entry><entry><title type="html">What goes in a CRA technical file, and how to keep it current</title><link href="https://xyanlabs.com/blog/what-goes-in-a-cra-technical-file/" rel="alternate" type="text/html" title="What goes in a CRA technical file, and how to keep it current" /><published>2026-10-24T00:00:00+01:00</published><updated>2026-10-24T07:50:58+01:00</updated><id>https://xyanlabs.com/blog/what-goes-in-a-cra-technical-file</id><content type="html" xml:base="https://xyanlabs.com/blog/what-goes-in-a-cra-technical-file/"><![CDATA[<p>The CRA’s technical documentation is set out in <code class="language-plaintext highlighter-rouge">Annex VII</code>. It has to exist before the product is placed on the market and be kept up to date for the support period (<code class="language-plaintext highlighter-rouge">Art. 31</code>). You keep it for ten years or the support period, whichever is longer (<code class="language-plaintext highlighter-rouge">Art. 13(13)</code>). This post is the first in a series on the technical file. Later parts take the risk assessment, the development process, vulnerability handling and the SBOM in turn.</p>

<h2 id="the-eight-parts">The eight parts</h2>

<ol>
  <li><strong>A general description</strong> of the product: its intended purpose, the software versions that affect compliance, photographs or illustrations for hardware, and the user information of <code class="language-plaintext highlighter-rouge">Annex II</code>.</li>
  <li><strong>Design, development, production and vulnerability handling</strong>: the system architecture and how the software components fit together, and the production and monitoring processes. For vulnerability handling: the SBOM, the coordinated vulnerability disclosure policy, the contact address for reports and how updates are distributed securely.</li>
  <li><strong>The cybersecurity risk assessment</strong>, including how each essential requirement in <code class="language-plaintext highlighter-rouge">Annex I Part I</code> applies to the product.</li>
  <li><strong>The information used to set the support period.</strong></li>
  <li><strong>The standards applied</strong>: harmonised standards, common specifications or certification schemes, and where none were applied, how each essential requirement is met instead.</li>
  <li><strong>Test reports</strong> that verify conformity with <code class="language-plaintext highlighter-rouge">Annex I Parts I and II</code>.</li>
  <li><strong>A copy of the EU declaration of conformity.</strong></li>
  <li><strong>The SBOM</strong>, where a market surveillance authority asks for it with a reasoned request.</li>
</ol>

<h2 id="why-a-file-goes-stale">Why a file goes stale</h2>

<p>Parts 1, 3 and 6 change every time the product does. The SBOM changes with every dependency bump. The risk assessment has to be updated when new information arrives (<code class="language-plaintext highlighter-rouge">Art. 13(3)</code>). A file written once as a set of PDFs is accurate on the day it is signed, and drifts from then on.</p>

<h2 id="a-living-file-with-its-processes-in-it">A living file, with its processes in it</h2>

<p>The file is more than a description of the product. Much of it is process: how the product is developed, how its vulnerabilities are handled, how a report under <code class="language-plaintext highlighter-rouge">Art. 14</code> gets made. The file has to show that those processes exist and that this product follows them.</p>

<p>Not every process needs writing for each product. Incident handling and reporting, supplier checks and secure development usually run across the whole company. The file can point to the company-wide procedure and name the version that applies. Beside it sits the product’s own evidence: the triage log, the reports sent, the reviews done. When the procedure changes, every product’s file picks up the change from one place.</p>

<h2 id="one-file-for-every-obligation-on-the-product">One file for every obligation on the product</h2>

<p>The CRA is rarely the only thing a connected product has to meet. The same product may be sold in the UK under PSTI, or made by a company whose ISO/IEC 27001 ISMS already sets out how it develops software securely. Much of the evidence is shared. A vulnerability disclosure policy serves the CRA and PSTI alike. The secure development process the ISMS defines can be the one the CRA file relies on.</p>

<p>So put all of it in one file, and map each obligation, whichever regime it comes from, to the same evidence. A change to a process then updates its evidence once, and an auditor for any regime finds what they need in the same place. Separate document sets for each regime drift apart.</p>

<h2 id="a-structure-that-holds-up">A structure that holds up</h2>

<p>We build files in three layers:</p>

<ul>
  <li><strong>Fundamentals.</strong> What changes rarely: the product definition, the regulatory scope and class, the development and vulnerability handling processes (or references to the company-wide ones), the support period.</li>
  <li><strong>A release record per version.</strong> The risk assessment as it stands, the traceability from each obligation to its evidence, the SBOM, the test reports and the declaration of conformity for that release.</li>
  <li><strong>Release history.</strong> Each release’s record, signed and kept, so you can show what was true for any version an authority asks about.</li>
</ul>

<p>Keep the requirements, threats and traceability as structured files in the same repository as the code. Then a release that changes a component also changes its evidence in the same pull request, and a check in CI can refuse a release whose record is incomplete.</p>

<p>If you’re starting a file, or inheriting one, our <a href="/services/#assessment">CRA assessment</a> checks each essential requirement against what the product already does. It ends with a plan for writing the rest.</p>]]></content><author><name></name></author><summary type="html"><![CDATA[Annex VII lists eight things the technical documentation must contain. The harder part is keeping it true for every release over a support period of five years or more.]]></summary></entry></feed>