Since 15 March 2026, a publicly trusted TLS certificate can be valid for at most 200 days. Until that date the ceiling was 398 days. The change comes from CA/Browser Forum ballot SC-081v3, adopted in April 2025, and it is only the first step: the limit falls to 100 days on 15 March 2027 and to 47 days on 15 March 2029. Eight months into the new regime, the pattern is clear. Websites with automated renewal did not notice anything. The outages came from the certificates nobody remembered: an appliance renewed by hand, a staging host on a forgotten subdomain, a wildcard copied onto six servers. This guide explains what changed, why, and how to get ready for the 100-day step before March 2027. It belongs to our guide to external attack surface management, because a certificate you renew is first of all an asset you have to know exists.
Key takeaways
- Maximum validity of a public TLS certificate: 200 days from 15 March 2026, 100 days from 15 March 2027, 47 days from 15 March 2029
- The period for which a CA can reuse your domain validation shrinks on the same calendar, down to 10 days in 2029
- The rules apply to publicly trusted certificates, not to those issued by your own internal CA
- Manual renewal does not scale: at 47 days you renew every certificate at least eight times a year
- The first step is an inventory of every host that presents a certificate, including the ones outside your CMDB
What the CA/Browser Forum decided
The CA/Browser Forum is the body where certificate authorities and browser makers agree the Baseline Requirements that every publicly trusted TLS certificate must meet. Ballot SC-081v3, "Introduce Schedule of Reducing Validity and Data Reuse Periods", was proposed by Apple and endorsed by Sectigo, Google Chrome and Mozilla. Voting closed on 11 April 2025: among certificate issuers there were 25 votes in favour, none against and 5 abstentions, and all four voting certificate consumers (the browser makers) voted yes. The ballot text sets a schedule from March 2026 to March 2029.
| From | Maximum certificate validity | Maximum reuse of domain validation |
|---|---|---|
| Until 14 March 2026 | 398 days | 398 days |
| 15 March 2026 | 200 days | 200 days |
| 15 March 2027 | 100 days | 100 days |
| 15 March 2029 | 47 days | 10 days |
Some CAs moved earlier and stay one day under the limit. DigiCert, for example, has issued certificates of at most 199 days since 24 February 2026, and plans 99 and 46 days for the next two steps. The schedule applies to certificates issued from each date: a 398-day certificate issued before 15 March 2026 stays valid until it expires, which is why many organisations met the first real renewal under the new rule only in the autumn.
The second column matters as much as the first. When a CA issues a certificate it checks that you control the domain. Until March 2026 that check could be reused for 398 days; in 2029 it will be valid for 10 days. In practice, at the last step every renewal also means a fresh domain validation, and that validation has to be automated as well.
Why the industry shortened certificate lifetimes
The ballot's own rationale is about revocation. When a private key leaks or a certificate is issued by mistake, the CA revokes it, and browsers are supposed to notice. The ballot describes relying on revocation as untenable for three reasons: status services do not scale with the volume of requests, they are not reliable, and checking them slows down page loads. Expiry, by contrast, is enforced by every client, every time. A shorter lifetime caps how long a compromised or misissued certificate can be abused, and keeps the information inside a certificate closer to reality. The ballot also names two side effects it wants: forcing automation, and making future cryptographic transitions faster. With post-quantum algorithms on the way, an organisation that can replace every certificate in a week is better placed than one that needs a project for it.
What breaks when lifetimes shrink
The web server with an ACME client renewing every 60 days was already fine under the old rules and is fine now. The trouble is everywhere else:
- Manual renewals tracked in a spreadsheet or a calendar. At 398 days a forgotten renewal happened once a year. At 200 days it happens twice, at 100 days four times, at 47 days at least eight times, for every certificate.
- Appliances and devices where the certificate is uploaded by hand: load balancers, SSL VPN portals, firewall management pages, web application firewalls, mail gateways, NAS and video systems with a public interface. Many have no ACME support, and the renewal procedure lives in the head of one person.
- Wildcards copied across servers. One certificate, installed on six machines by three teams. Renewing it means finding all six.
- Certificates managed by suppliers. A marketing platform, a booking engine or a support portal on your subdomain, with a certificate renewed by the vendor. If the vendor fails, the error shows your name.
- Applications that pin a certificate instead of trusting the CA chain. Every renewal breaks them unless they are updated at the same time.
- Incomplete chains after renewal. A renewed certificate installed without the right intermediate works in one browser and fails in another, or in an API client.
Internal CAs are out of scope, with one trap
The Baseline Requirements govern publicly trusted certificates. Certificates issued by your own internal CA, for example by Active Directory Certificate Services for intranet servers, are not bound by the ballot and can keep their current lifetime. The trap is the opposite case: internal systems that use a public certificate because it was easier, such as an intranet with a public wildcard. Those follow the new rules, and they are often the ones managed by hand.
Inventory first: you cannot renew what you do not know exists
Every renewal plan starts from a list, and the list is usually incomplete. Over the years certificates are issued for test environments, campaign microsites, old portals, supplier-hosted services and hosts created by a developer for a demo. The CMDB knows some of them. The internet knows all of them: every publicly trusted certificate is logged in public Certificate Transparency logs, and anyone, an attacker included, can list the names your organisation has requested certificates for.
A useful inventory answers five questions for each certificate:
- Which host or hosts present it, and on which ports
- Who issued it and when it expires
- How it is renewed: ACME, vendor API, supplier, or by hand
- Who owns it: a named person or team, not "IT"
- Whether the host still needs to exist
The last question is the most valuable one. Every forgotten host you switch off is a certificate you no longer have to renew and a piece of shadow IT an attacker can no longer use. Shorter lifetimes are a good reason to reduce the attack surface rather than only automate it.
Automating with ACME, and where it fails
ACME is the protocol, standardised as RFC 8555, that lets a server request, validate and install a certificate without a person in the loop. Most commercial CAs support it as well as Let's Encrypt, and most web servers, reverse proxies, Kubernetes ingress controllers and hosting panels have an ACME client. For these, the job is to switch it on, set renewal well before expiry and monitor that it actually runs.
Two technical choices matter as validation reuse shrinks:
- HTTP-01 validation works when the host is reachable from the internet on port 80. It is simple, but not suitable for internal hosts or wildcards.
- DNS-01 validation proves control by publishing a TXT record. It works for wildcards and for hosts that are not reachable, but it needs a DNS provider with an API and a narrowly scoped credential to write records. Treat that credential like a secret: whoever holds it can obtain certificates for your domain.
Where ACME does not reach, you have three options: the vendor's own API or management tool for the appliance, a certificate lifecycle management platform that pushes certificates to devices, or terminating TLS in front of the device on a component that does support automation, such as a reverse proxy or load balancer. The worst option is the current one: a person, a reminder and a procedure nobody has run for a year.
A monitoring routine that catches what automation misses
Automation fails quietly: an expired DNS API token, a firewall rule that blocks validation, a renewed certificate that the service never reloads. Monitoring is what turns a silent failure into a ticket.
- Measure from the outside. Check the certificate each host actually serves, not the file on disk. The reload step is where renewals often fail.
- Alert in stages: for example at 30, 14 and 7 days before expiry, sent to the certificate owner, not to a shared mailbox.
- Treat "renewed but not served" as an incident. If the expiry date seen from outside does not move after the renewal job ran, something in the chain is broken.
- Check the chain and the protocol. A complete chain and current TLS versions belong to the same check.
- Watch for new names. A new subdomain that appears in Certificate Transparency or in DNS is an asset that needs an owner and a renewal path from day one.
Checklist for 15 March 2027 (100 days)
- Build the inventory of every host that presents a public certificate, starting from your domains, not from your CMDB.
- Switch off what is no longer needed, and record the decision.
- Give every certificate a named owner and a documented renewal method.
- Move every web server, proxy and ingress to ACME, with renewal at about two thirds of the lifetime.
- For appliances without ACME, choose vendor API, lifecycle tooling or TLS termination in front, and test it before March.
- Move domain validation to DNS-01 where needed, with a scoped DNS API credential.
- Ask suppliers that serve your subdomains how they renew, and put it in the contract.
- Set staged expiry alerts measured from outside, and test them with a short-lived certificate.
- Find applications that pin certificates and change them to trust the CA chain or pin the public key you control.
- Plan for 2029: if a step in your process cannot run eight times a year without a person, it is not ready for 47 days.
How Orizon helps
The hard part of shorter lifetimes is not renewing the certificates you know about; it is finding the ones you do not. RECON External starts from your domain name and finds every subdomain, forgotten test hosts included, then shows the issuer, validity and expiry of the certificate on every host, so a certificate never lapses unnoticed. Scans can be scheduled daily, weekly or monthly, and the timeline shows what changed since the previous one, including new hosts that appeared. RECON does not renew certificates: it gives you the complete list and the warning in time. To build that inventory as a continuous process rather than a one-off exercise, see our attack surface management approach.