A failed DNSSEC key rollover in Albania's .al top-level domain left the zone's domains unresolvable for validating resolvers on July 3, 2026, and prompted 1.1.1.1 to make a long-invisible mitigation visible to clients. Cloudflare's public resolver installed a Negative Trust Anchor to restore resolution and, for the first time, attached a new Extended DNS Error code to every response served under it.
The .al TLD is operated by AKEP, the Albanian communications authority, and hosts government services, banks and media; it sits at #191 on Cloudflare Radar's TLD ranking. The breakage threatened every .al domain regardless of hosting or authoritative nameservers. It came two months after a comparable incident in .de, where the same NTA remedy was used.
The chain of trust and where it snapped
DNSSEC links the root zone to child zones through Delegation Signer (DS) records, each holding a fingerprint of the child's DNSKEY. A resolver validating .al compares the DNSKEY its nameservers return against the DS record in the root; a match lets the resolver trust responses from those nameservers, and the pattern repeats for signed child zones below. Any break in that chain, such as a DS record pointing at a key that no longer exists, invalidates everything beneath it.
Before July 3, the root held a DS record matching the DNSKEY served by .al:

The sequence of operator actions then unfolded in three steps, each leaving validation broken until the last one:
- Around 14:15 UTC, the operator published a new DNSKEY and stopped serving the old one, while the root DS record still referenced the old key (id=26319). Resolvers found no matching key and failed.
- Around 17:00 UTC, the new DNSKEY was withdrawn without restoring the old one, leaving the zone with no DNSKEY records at all while the DS record still pointed at id=26319. Resolution kept failing.
- Around 19:15 UTC, the DS record was removed from the root. With no DS record, resolvers no longer expected DNSSEC validation for
.al, and resolution returned — with the entire TLD unsigned.

The zone remains unsigned as of publication; the DS record has not been restored, so no .al domain can use DNSSEC protections.

Restoring service with an NTA
A Negative Trust Anchor, specified in RFC 7646, instructs a recursive resolver to treat a zone as unsigned and skip validation. Cloudflare attempted to contact the .al operator directly and raised the problem on the DNS-OARC Mattermost, but got no response — the operator's contact addresses were themselves under .al and unreachable during the outage. The NTA reached all 1.1.1.1 users by 17:15 UTC, about three hours after the chain broke.

The tradeoff mirrors the .de case: while the NTA is active, .al domains are unprotected against DNS spoofing. Cloudflare judged that acceptable because the failure was public, confirmed and affecting every validating resolver alike. The NTA was removed the next day, once the DS record had been pulled from the root and validation was no longer expected.
What made the remedy opaque is that, until now, an NTA-served response was indistinguishable from a fully validated one. RFC 7646 acknowledges the gap and recommends out-of-band public disclosure of installed NTAs; status pages were published for both incidents, but a status page has to be sought out. An application, a monitoring tool or a user querying 1.1.1.1 had no way to learn from the response itself that validation had been bypassed.
Signalling the bypass in-band
Extended DNS Error codes (RFC 8914) let resolvers attach context to any response, successful or not. Babak Farrokhi at Quad9 proposed an Internet-Draft, "Disclosure of Negative Trust Anchors in DNS Responses," adding an EDE code for exactly this situation; Cloudflare co-authored it, and 1.1.1.1 now implements it. During the incident, any query for an .al name returned the answer together with the new code:
$ kdig @1.1.1.1 google.al
;; ->>HEADER<<- opcode: QUERY; status: NOERROR; id: 32848
;; Flags: qr rd ra; QUERY: 1; ANSWER: 1; AUTHORITY: 0; ADDITIONAL: 1
;; EDNS PSEUDOSECTION:
;; Version: 0; flags: ; UDP size: 1232 B; ext-rcode: NOERROR
;; EDE: 9 (DNSKEY Missing): 'no SEP matching the DS found for al.'
;; EDE: 33 (Negative Trust Anchor): 'a Negative Trust Anchor has been applied for this query (see RFC 7646)'
;; ANSWER SECTION:
google.al. 300 IN A 142.251.142.196
A google.al lookup came back as NOERROR with a valid answer but two EDE codes attached. EDE 9 (DNSKEY Missing) exposes the underlying failure — the chain of trust was broken and validation failed — while EDE 33 (Negative Trust Anchor) indicates that 1.1.1.1 applied an NTA and served the response anyway. Read together, they tell a client that the answer is genuine but not cryptographically validated.
EDE 33 is emitted on any response generated while an NTA is active, whether or not the queried name would have failed validation — an unsigned domain under an active NTA still gets the code. That is deliberate, since the NTA covers the whole zone. It also fixes a problem noted in the .de write-up, where 1.1.1.1 returned EDE 22 (No Reachable Authority) instead of the underlying DNSSEC error; during the .al incident, EDE 9 (DNSKEY Missing) came back correctly alongside EDE 33.
EDE 33 has been assigned by IANA. The draft is an individual submission before the IETF DNSOP Working Group; Knot's kdig now recognises EDE 33 by name, a pull request adding support to Unbound is under review, and the working group is scheduled to discuss the draft at the IETF meeting in Vienna, July 18–24.
Why it matters
TLD-level DNSSEC failures are rare, but a single one takes out every domain beneath it at once and hits every validating resolver the same way. Two such incidents in two months show that NTAs remain a necessary operational tool even though, until now, their use was invisible to the people they affected. EDE 33 closes the gap RFC 7646 left open: a response served under an NTA now says so, giving operators, monitoring tools and users what they need to understand what the resolver did and why.



