security-txt.eu

AdviceAs of 9 min read

Setting up a security contact: address, mailbox, ownership

Writing a security.txt takes minutes. What matters is what happens behind the address in the file: who reads the mailbox, who covers during holidays, who replies and by when? This page shows how to set up a security contact that still works on a Friday evening in August.

Why the contact matters more than the file

Someone who finds a vulnerability in your website, your software or your device looks for a way to tell you. If they find none, or never hear back, there are three typical outcomes: the report gets lost, it goes to a national CSIRT instead, or the issue is published without coordination. None of these is good for you.

security.txt only solves the first part, being found. It states in a machine-readable way where reports should go. Whether anyone reads, understands and answers them is up to your organisation. A checker can tell you whether the file is valid. It cannot tell you whether the mailbox behind it is alive. That is where the most common failure lies.

Which address belongs in security.txt

Besides Expires, Contact is the only required field (RFC 9116 section 2.5.3). It may appear more than once, listed in order of preference. For email, RFC 9116 points to the convention in RFC 2142: security@ your domain. It is where anyone reporting a security issue looks first, file or no file.

AddressSuitable forVerdict
security@companies without their own products, websites, service providersThe standard under RFC 9116 and RFC 2142. Enough for most organisations.
psirt@ and csirt@manufacturers of software, devices, machines with softwareRequired by the German BSI TR-03183-3: psirt@ for your products, csirt@ for your own IT (4.3.1 f to i).
info@, contact@nobodyReports end up between sales enquiries, job applications and invoices, and everyone reads the spam.
firstname.lastname@nobodyFails during holidays, sick leave and when people leave. Under the TR the address must show its function (4.3.1 g and i).
Free mail addressnobodyLooks unprofessional, sits outside your control and cannot be proven to be yours.

The address must be on your own domain and spelled the same way in the file, in your policy and on your reporting page. If you follow the TR, list the PSIRT mailbox first in security.txt, then the CSIRT mailbox, then the reporting page (4.2.3). All other fields are covered in the guide.

A shared mailbox, not a forward to one person

The address should lead to a mailbox of its own that several people can access. The TR requires exactly that: functional mailboxes for both roles, with access for everyone in those roles (4.3.1 c). Three setups are common:

  • Shared mailbox (a shared mailbox in Microsoft 365, a group mailbox elsewhere). Everyone sees the same messages and the same read status. The best option, because it stays visible what has been handled.
  • Distribution list forwarding to several people. Works, but everyone has their own copy. Nobody knows whether someone else is already replying, and in the end two people answer, or nobody does.
  • Forward to one person. Fails as soon as that person is away. Out-of-office replies also hand address harvesters a name and a deputy.

Check your provider's spam filter too. Reports often come from unknown senders, with attachments or code in the body, exactly what filters like to discard. For the security mailbox: flagging and sorting into folders, yes; rejecting or deleting unseen, no. How to do that with rules is described under security.txt spam.

Who reads it: roles, cover, availability

A mailbox without named readers is not a contact. Write down who reads the mailbox, who decides and who stands in. For manufacturers, the TR explicitly requires defining the tasks and authorisations of each role (4.3.1 o and p) and spreading the roles across several people, unless the company is a microenterprise (4.3.1 b). There should be enough people to cover absences, so that response times and commitments still hold (4.3.1 m).

For a mid-sized company, this line-up is usually enough:

RoleTaskTypically
First readergoes through the mailbox every working day, sorts, confirms receiptIT administration, managed service provider
Deputytakes over during holidays and sick leave, double-checks every classificationa second person from IT or development
Technical assessmentchecks whether the report holds up and what is affecteddevelopment, product owners, web agency
Decisiondecides on the fix, on informing customers and, for manufacturers, on notifying the authoritiesmanagement or a named lead

Two things are often overlooked. First, weekends and public holidays: a report arriving on Friday evening that mentions an ongoing attack should not wait until Tuesday. Agree whether and how someone can be reached outside working hours. Second, language: in a European company the mailbox may receive reports in several languages, but most come in English. Under the TR, both roles must at least be able to receive and process reports in English (4.3.1 n).

From report to reply

A fixed process keeps a report from falling between two owners. The deadlines below come from TR-03183-3 and are binding only for manufacturers that follow it. As a yardstick they work for any company.

StepWhat happensDeadline
1. Receiptfirst reader reviews, opens a case, notes time and senderevery working day
2. Triagevulnerability report, beg bounty, advertising or phishing? Signs of active exploitation go straight to the decision makersame day
3. First replya personal reply to the reporter, not an automated onewithin 5 working days (TR 4.4.8 a)
4. Detailed feedbackconfirmed or rejected, specific questions, or an explanation of why it takes longerwithin 10 working days (TR 4.4.8 b and c)
5. Fix and disclosureremediation, informing those affected, publicationconfirmed issues disclosed within 90 days (TR 4.4.10 a)
6. Closureclose the case, inform the reporter, file the recordwithout undue delay (TR 4.4.11 b)

Two TR rules deserve special attention. No case may be closed by a single analyst, so that no valid vulnerability is dismissed as noise (4.4.7 g). And anonymous reports are exempt from the response times, since there is nobody to reply to (4.4.8 a). Deadlines and commitments belong in your coordinated vulnerability disclosure policy; the CVD template drafts one with all required content.

The TR does not rule out an automatic acknowledgement, but it does not count as the first reply, which must explicitly not be automated (4.4.8 a). Keep it short and link to your policy. Otherwise it mainly confirms to address harvesters that the mailbox is read.

Encryption and a web form

Someone reporting a serious flaw does not want to send it across the internet in plain text. RFC 9116 therefore recommends encryption for email contacts (section 2.5.3). The TR makes it mandatory: each functional mailbox gets its own OpenPGP key, the .asc file is available for direct download, and its fingerprint is published next to it (4.3.1 j, 4.3.2 a and b). Keys may be valid for at most five years (4.3.1 l). Creating, storing and renewing keys is covered in signing security.txt with GnuPG.

Access is the part people forget: the private key must be available to everyone who reads the mailbox, including the deputy. An encrypted report that nobody can open while a colleague is on holiday is as good as no report.

A web form is a useful addition to the address. Required fields for product, version, steps to reproduce and observed impact produce more complete reports than free-form email. The TR requires manufacturers to offer such a form on their own website, usable anonymously and at least in English (4.3.3 a to d), plus a central reporting page that is fully readable without JavaScript and without a login (4.5.1 d). Our own reporting page shows one way to do it: two mailboxes with keys, fingerprints, a form and a policy.

Testing that reports get through

The simplest test is also the most important one: from an external address, such as a personal mailbox, send an email to every address in your security.txt. Does it arrive? In the inbox or in spam? Does anyone notice without being told? How long until the first reaction?

The TR recommends writing to the functional mailboxes from an external address at least daily, ideally automatically (4.3.1 q), and submitting the web form daily as well (4.3.3 f). For a smaller company, a quarterly test with a real human response is often more revealing than a daily robot whose results nobody reads. The TR requires reviewing the contents of security.txt at least every quarter anyway (4.2.9 c). Put both on the same date.

Check the file while you are at it: are addresses and keys still correct, is the expiry date far enough away? The checker does that in seconds.

Keeping records

For every report, record when it arrived, who read it, how it was classified, when you replied and how the case ended. It helps with follow-up questions from the reporter and with the next similar report. For manufacturers under the Cyber Resilience Act, evidence that a contact address for vulnerability reports is in place is even part of the technical documentation (CRA Annex VII point 2(b)). A spreadsheet or a ticket system is enough, as long as everyone involved can access it.

What the Cyber Resilience Act requires from manufacturers

For manufacturers of products with digital elements, meaning software and devices with software, the EU Cyber Resilience Act (Regulation (EU) 2024/2847) turns good practice into an obligation:

  • Single point of contact. Manufacturers designate a single point of contact that lets users communicate with them directly and rapidly, including to report vulnerabilities. It must be easy to identify and must not be limited to automated tools (Article 13(17)).
  • Contact address for vulnerabilities. The essential requirements include a contact address for reporting vulnerabilities in the product (Annex I Part II point 6) and a coordinated vulnerability disclosure policy (point 5). These apply from 11 December 2027 (Article 71(2)).
  • Reporting obligation. Since 11 September 2026, manufacturers must report every actively exploited vulnerability in their product they become aware of: early warning within 24 hours, notification within 72 hours, final report no later than 14 days after a fix is available (Article 14(2), Article 71(2)). This also covers products placed on the market before December 2027 (Article 69(3)).

The CRA prescribes no format, not even security.txt. For the mailbox, though, the reporting obligation means that an email describing an actively exploited flaw may start a 24-hour clock. Looking at the mailbox once a week is not enough then. Whether and from when your company is affected depends on the individual case; this page is not legal advice.

In addition to your own reading. The CRA Response Center reads the security mailbox alongside you, assesses every email and alerts the right people by email and text message when an email may matter. It is an additional alarm system: it replaces neither the mailbox nor your own reading, and it never submits reports. It connects through a simple forward, no access, no password. The AI runs in Germany; your emails never go to OpenAI, Anthropic or Google. Visit the CRA Response Center

Frequently asked questions

Which email address should I use for security reports?

security@ your domain, as RFC 9116 recommends by reference to RFC 2142. Manufacturers following BSI TR-03183-3 set up psirt@ for products and csirt@ for their own IT. Never info@, never a personal address, never a free mail account.

Is forwarding security@ to our IT manager enough?

Only while they are in the office. A shared mailbox with at least two readers is better: holidays and sick leave leave no gap, and everyone can see who is already handling a report.

How quickly do I have to reply to a vulnerability report?

RFC 9116 sets no deadline. BSI TR-03183-3 requires a personal first reply within 5 working days and detailed feedback within 10 working days. Manufacturers under the CRA must also send an early warning within 24 hours for actively exploited vulnerabilities.

Can I send an automatic acknowledgement?

Yes, but it does not count as a reply under the TR, which requires a personal one. Keep it short, without names, and link to your policy, or it mostly helps address harvesters.

What is a PSIRT and do I need one?

A Product Security Incident Response Team handles vulnerabilities in a company's own products. You do not need it as a department, but you do need the function: named people, a mailbox and a process. The TR requires the role for manufacturers that follow it.

Does the security mailbox have to accept encrypted email?

RFC 9116 recommends it; the TR requires a dedicated OpenPGP key for each functional mailbox. Everyone who reads the mailbox needs the private key, including the deputy.

How do I test whether my security contact is reachable?

Send an email from an external address to every address in your security.txt and see whether it arrives and whether anyone reacts. The TR recommends a daily, ideally automated test. The checker tests the file itself.

Sources