· via dev.to (home feed)
Post-quantum key exchange is everywhere; ML-DSA signatures are only now shipping
Hybrid ML-KEM key exchange quietly became the default in major browsers and tools, while post-quantum signatures are only now reaching identity software, starting with authentik's ML-DSA release.

The migration is split in two
Open a TLS handshake on a major website in a current browser and the session key was most likely agreed with X25519MLKEM768: a classical elliptic-curve exchange paired with ML-KEM-768, the lattice-based key encapsulation mechanism NIST standardized in FIPS 203. According to a dev.to analysis, Chrome, Firefox, OpenSSL and OpenSSH enabled this without any announcement, and the internet kept working.
That half of the post-quantum migration moved first for a structural reason, the post explains. Key exchange is retroactively attackable: an adversary can record encrypted traffic today and decrypt it once a sufficiently large quantum computer exists. NIST's transition report, IR 8547, published October 27, 2026, treats key exchange as the urgent half for exactly this reason.
Signatures are the second half, and they are only now reaching shipping software.
Size is why signatures lag
A signature does different work: it proves a certificate, token, passkey assertion or software update came from whom it claims. A forgery only matters while the verifier still trusts the key, so there is no retroactive threat pushing replacement. The practical blocker is bulk. An Ed25519 signature is 64 bytes; an ML-DSA-65 signature runs about 3.3 kilobytes with a public key near 2 kilobytes. Putting that into every certificate chain, JWT and passkey assertion bloats handshakes and tokens beyond what the carrying protocols were designed for.
The result, per the dev.to piece: two years after NIST finalized ML-DSA in FIPS 204, no browser-trusted certificate authority issues ML-DSA certificates and no browser accepts one. The standards plumbing exists — RFC 9881, from October 2025, defines ML-DSA in X.509 certificates, and RFC 9935, from March 2026, covers ML-KEM keys — but public trust has not caught up.
What authentik shipped
On November 3, the open-source identity provider authentik released version 2026.11, which the author calls the clearest example of the signature half moving from roadmap to release. It adds ML-DSA in three places:
- Certificates. authentik can generate and store ML-DSA-44, ML-DSA-65 and ML-DSA-87 certificates alongside RSA, ECDSA and Ed25519. These target areas the operator controls: internal TLS, mutual TLS against a private CA, and RADIUS with EAP-TLS 1.3 — not public login pages, since public CAs and browsers do not accept them yet.
- OIDC tokens. Providers can sign ID tokens, access tokens and client assertions with an ML-DSA certificate, and proxy outposts verify ML-DSA-signed tokens from upstream identity providers. The caveat authentik states plainly: applications that validate tokens themselves need a JWT library that understands ML-DSA before switching a provider over.
- Passkeys. WebAuthn registration and login accept ML-DSA credentials. No shipping authenticator supports ML-DSA yet, and some browsers error on unknown algorithms, so authentik puts it last in the accepted list — support in place and tested before the hardware arrives.
The same release raises SAML's default symmetric encryption to AES-256-GCM. AES-128 is not broken by quantum computers, but Grover's algorithm erodes its margin and the NSA's CNSA 2.0 suite requires AES-256 for national security systems. authentik is equally open about what is blocked: ML-DSA key generation fails under its FIPS compliance mode because the validated OpenSSL FIPS provider does not implement ML-DSA; SAML signing waits on an XML Signature algorithm identifier; ML-KEM token encryption waits on JOSE and XML Encryption definitions.
A candidate answer for public certificates
The size problem has one serious proposal at internet scale. On September 29, Cloudflare announced it intends to become a public certificate authority, acquiring trusted root key material from GlobalSign, with production issuance of Merkle Tree Certificates scheduled for the first quarter of 2027. Instead of signing each certificate individually with a large post-quantum signature, the CA signs a batch once as a Merkle tree, and servers present an inclusion proof. The draft, draft-ietf-plants-merkle-tree-certs, was at revision 08 on November 2, and Cloudflare's applications to the Chrome, Apple, Microsoft and Mozilla root programs remain pending — but the dev.to author calls it the first concrete plan for public post-quantum certificates, with a date next quarter rather than next decade.
What operators should do now
The deadlines are fixed. NIST IR 8547 deprecates RSA, ECDSA, EdDSA and elliptic-curve Diffie-Hellman after 2030 and disallows them after 2035. The EU roadmap asks member states to start national transition plans by the end of 2026. CISA's January 2026 product categories list marks browsers, web servers and cloud platforms as widely available with post-quantum capability — meaning federal buyers should only buy capable products in those categories now — while identity software sits in a transitioning table where vendors are encouraged to implement and test immediately. The unglamorous work for identity operators: inventory every certificate and signing key by algorithm.
Why it matters
The asymmetry in the post-quantum migration is now visible in shipping software: the confidentiality half is largely done, the authentication half is just starting. Site operators cannot obtain browser-trusted ML-DSA certificates yet, but they can deploy ML-DSA where they control both ends — internal TLS, private CAs, token signing — and check that JWT libraries and authenticators will handle it. The author's own experience adding ML-DSA-65 identity signatures to a small end-to-end encrypted messenger points to the same rule of thumb: a 3.3-kilobyte signature that travels once per session is cheap, while one per message is a tax on every send. Preparing now, before the 2030 deprecation date, turns a forced scramble into routine maintenance.
- #post-quantum
- #tls
- #certificates
- #identity
- #cryptography