The clause nobody enforced is about to disappear
Go find the sentence in the Baseline Requirements that tells a CA to think twice before it issues a certificate for yourbrand-account-verify.com. It's there. It has been there in some form since at least 2012. On December 2 it's gone.
Ballot SC105 strips the "High Risk Certificate Request" language out of the TLS Baseline Requirements and the EV Guidelines. Voting closed on October 28, 2026. Nobody voted against it and nobody abstained. The published tally says 27 Certificate Issuers voted yes, although the list of names printed under it only has 25. On the browser side Apple and Mozilla voted yes. Google and Microsoft aren't in the record at all. The IPR review period ends on December 2, 2026 at 11:00 UTC, and if nobody files an exclusion notice by then, the text is deleted.
Most people will read this as housekeeping. Fair enough, the rule was dead weight. But it was the one place in the BRs that told anybody to look at what a name says. Every other check asks who controls it. After December 2, reading the name is your job.
What SC105 actually removes
Stephen Davidson of DigiCert proposed the ballot, and Amazon Trust Services and SwissSign endorsed it. The redline is short. Five things go.
First, the definition. In full: "A Request that the CA flags for additional scrutiny by reference to internal criteria and databases maintained by the CA, which may include names at higher risk for phishing or other fraudulent usage, names contained in previously rejected certificate requests or revoked Certificates, names listed on the Miller Smiles phishing list or the Google Safe Browsing list, or names that the CA identifies using its own risk-mitigation criteria."
Second, the obligation in section 4.2.1. The CA "SHALL develop, maintain, and implement documented procedures that identify and require additional verification activity for High Risk Certificate Requests prior to the Certificate's approval."
Third, the clause that held Delegated Third Parties to the same standard.
Fourth, EV Guidelines section 3.2.2.12.1, which applied all of this to EV. The same ballot also cuts a line from the mixed-character-set rule that made EV CAs "visually compare any Domain Names with mixed character sets with known high risk domains." That was the homograph check.
Fifth, and the ballot summary never mentions this one: revocation reason 10 in section 4.9.1.1. It made a CA revoke within five days when "a Wildcard Certificate has been used to authenticate a fraudulently misleading subordinate Fully-Qualified Domain Name." The new text, in its entirety: "10. Removed;".
The ballot is right that the rule did nothing
Nothing of value is being lost. Here's why. The definition said "may include". A CA could write a high-risk procedure that checked nothing at all and still comply, because the BRs never said what had to be on the list. Auditors had nothing to test against. The ballot says so itself: the rule "does not define what constitutes a high risk request" and "gives CAs and auditors no objective basis for assessing compliance."
The biggest CA gave up on it years ago. Let's Encrypt checked domains against Google Safe Browsing when it launched, and its 2015 post on the CA's role in fighting phishing already argued that "CAs are not well positioned to operate anti-phishing and anti-malware operations." A note at the top of that post says that as of January 10, 2019 it no longer checks Safe Browsing. Josh Aas had it right. A CA sees a name and a challenge response. It never sees the page, and the page can change the day after issuance.
So don't mourn the clause. The argument for why it's safe to drop is a different matter.
The replacements answer a different question
The ballot says the High Risk language "has been overtaken by the concrete controls added to the BRs in recent years, including CAA processing, DNSSEC validation, Multi-Perspective Issuance Corroboration, pre-issuance linting, and the sunset of weaker validation methods."
Go through what each of those does. CAA lets a domain owner say which CA may issue for it. DNSSEC validation stops a forged DNS answer from passing a challenge. MPIC checks the challenge from several networks, so a local BGP hijack can't fake it. Linting catches malformed certificates. Retiring the weak validation methods closes holes in how you prove control.
All of them make sure the requester really controls the domain in the request. That's a real gain. It protects yourbrand.com from being issued to someone else.
Not one of them touches yourbrand-account-verify.com. The attacker registered it this morning, so they control it. They can publish CAA, sign the zone and pass the challenge from any vantage point you like. By every measure the BRs still use, the request is legitimate. Nobody moved the question "is this name built to fool someone?" to a better control. It just stopped being asked.
The wildcard reason went with it
Reason 10 gets its own section because it covered the trick CT monitoring is worst at.
A phishing operator registers example-cdn.top, gets a certificate for *.example-cdn.top, and serves the kit at login.yourbrand.com.example-cdn.top. People read hostnames left to right. The first thing this one tells them is your brand.
The certificate in the CT log says *.example-cdn.top. Your brand isn't in it anywhere. Nothing you search for will ever match it, because the misleading part of the hostname never went to the CA.
Until December 2, a CA that heard about this had a named revocation reason and a five-day clock. After that, the closest hooks left are reason 7, "The CA obtains evidence that the Certificate was misused," and reason 15, revocation required by the CA's own CP or CPS. Both survive. Both depend on the CA agreeing that a phishing subdomain counts as misuse. Some CAs will. Others will hand you the Let's Encrypt line: we validated the domain, the content isn't ours. We wrote about the blast radius of wildcards when the key is yours. This is what a wildcard looks like when the key belongs to the attacker.
What your CT watch can still see
SC105 doesn't touch Certificate Transparency. Every publicly trusted certificate still gets logged, and the logs are still your best early warning for lookalikes that put your brand in the registered name. The full setup is in using CT logs to catch phishing domains. Two quick checks give you a baseline today.
Search crt.sh for unexpired certificates whose names start with your brand:
curl -s "https://crt.sh/?q=yourbrand%25&output=json&exclude=expired" \
| jq -r '.[].name_value' | sort -u
The %25 is a URL-encoded %, which is crt.sh's wildcard. Only put it at the end. When we tried a leading wildcard (%25yourbrand%25), crt.sh sent back an empty array with no error. That's worse than a failure, because it looks like good news. A prefix search finds yourbrand-login.com and yourbrand.evil.site. It misses secure-yourbrand.com.
For lookalikes that don't start with your brand, generate permutations and check which ones have live certificates:
dnstwist --registered --format json yourbrand.com | jq -r '.[].domain' > lookalikes.txt
while read -r d; do
curl -s "https://crt.sh/?q=${d}&output=json&exclude=expired" \
| jq -r --arg d "$d" 'if length > 0 then "\($d): \(length) unexpired certs" else empty end'
sleep 2
done < lookalikes.txt
Use the JSON output. dnstwist's source prints --format list and returns before it does any DNS lookups, so --registered quietly does nothing there and you end up hammering crt.sh with thousands of names nobody registered. Leave the sleep in, too. crt.sh is a free public service, and it's slow enough already.
Don't mistake this for monitoring. It's a one-off baseline. For something that runs all the time, read the logs directly. More and more of them serve the static, tiled API now, which we covered in Certificate Transparency went static. And if you run certificates for clients, check their brands too. They won't.
Catch the phish that CT will never show you
The wildcard-subdomain case needs a signal that has nothing to do with the certificate. The cheapest one is already sitting in your logs. Lazy phishing kits hotlink the real site's CSS, logo and fonts, so every victim's browser fetches your assets with a Referer that points at the phishing host.
Give static assets a log of their own in nginx:
log_format asset_referers '$time_iso8601 $http_referer $uri';
server {
# ...
location ~* \.(css|js|svg|png|woff2)$ {
access_log /var/log/nginx/access.log combined;
access_log /var/log/nginx/asset-referers.log asset_referers;
}
}
The log_format line belongs in the http block. You need both access_log lines, because an access_log inside a location replaces the inherited ones. It doesn't add to them. Then list every referring host that isn't yours:
awk '{print $2}' /var/log/nginx/asset-referers.log \
| grep -vE '^(-|https://(www\.)?yourbrand\.com/)' \
| awk -F/ '{print $3}' | sort | uniq -c | sort -rn | head -20
You'll get noise: translation proxies, archive services, the partner who embeds your logo. Something like login.yourbrand.com.example-cdn.top won't need explaining. Careful kits copy the assets and set no-referrer, and this misses them. It costs nothing, though, and it covers the one case your CT watch can't.
Report to the people who can block it
You can still file a Certificate Problem Report after December 2. Section 4.9.3 still makes every CA publish instructions for reporting "Certificate misuse, or other types of fraud" in section 1.5.2 of its CPS. Section 4.9.5 still gives the CA 24 hours to investigate and send a preliminary report. So file one. Just don't expect much from it.
Revocation was always a weak weapon against phishing. When GoDaddy revoked 378,067 certificates in October, Chrome kept loading them for weeks. A Safe Browsing listing puts a full-page red warning in front of Chrome, Firefox and Safari users, and a registrar can take the domain away entirely. Report the phishing URL to Google Safe Browsing first. Then the registrar's abuse contact. The CA comes last.
One more ballot to keep an eye on. SC107, "Improve Certificate Problem Reports and Clarify the Meaning of Revocation," is in its discussion period. Nobody has voted on it, so none of it binds anyone yet. If it ends up defining what "misused" means, part of the wildcard gap could close again. Don't plan around it until it passes.
Frequently asked questions
Does SC105 change anything about how my own certificates are issued? For a legitimate subscriber, almost nothing. Your CA still has to validate domain control, check CAA and run multi-perspective validation. A CA that used to put certain names through an extra manual check may stop doing that, but SC105 explicitly leaves CAs free to keep risk-based controls in their own CP/CPS, and some will.
Will CAs start issuing certificates they used to refuse? For DV you'll barely notice, since Let's Encrypt, the largest issuer, hasn't screened names against Safe Browsing since January 2019. The bigger change is in EV, where the Guidelines explicitly required a visual comparison of mixed-script names against known high-risk domains, and that requirement is gone.
Can I still get a CA to revoke a certificate used for phishing? You can ask, and the CA has to investigate a Certificate Problem Report within 24 hours. Without a specific rule for wildcard abuse, revocation now rests on the general "misused" reason or on the CA's own policy, so results will vary from CA to CA. Report to Safe Browsing and the registrar as well, because they can take the page down faster than revocation can.
Does CAA protect me from lookalike domains?
No. CAA only covers the domain it's published on. It stops an unauthorised CA from issuing for yourbrand.com, but whoever registers yourbrand-support.com controls that domain's DNS and can publish any CAA record they want.
When exactly does the change take effect? The IPR review period for SC105 runs from November 2 to December 2, 2026 at 11:00 UTC. If no member files an exclusion notice before then, the updated Baseline Requirements and EV Guidelines apply from that moment.