NIST's post-quantum cryptography standards
Post-quantum cryptography · facts checked 2026-10-04
The facts
- In August 2024 NIST finalised its first three post-quantum cryptography standards, FIPS 203 for general encryption, and FIPS 204 and FIPS 205 for digital signatures. (NIST)
- FIPS 203 specifies ML-KEM, formerly CRYSTALS-Kyber; FIPS 204 specifies ML-DSA, formerly CRYSTALS-Dilithium; and FIPS 205 specifies SLH-DSA, formerly SPHINCS+. (NIST)
- When it released the standards, NIST urged system administrators to start building them into their systems straight away, because full integration takes time. (NIST)
- In March 2025 NIST chose HQC, which rests on error-correcting codes rather than lattices, as a backup to ML-KEM, and said it expected a final HQC standard in 2027. (NIST)
- NIST warns of harvest now, decrypt later, in which adversaries gather encrypted data today in order to decrypt it once quantum computing matures. (NIST)
- NIST IR 8547, a draft first issued in November 2024, proposes deprecating quantum-vulnerable algorithms such as RSA and elliptic-curve schemes at 112-bit strength after 2030 and disallowing them after 2035. (NIST)
The panel
Jordan Pike AI host
NIST finalised its first three post-quantum cryptography standards in August 2024: FIPS 203 for general encryption, and FIPS 204 and FIPS 205 for digital signatures. NIST says FIPS 203 specifies ML-KEM, formerly CRYSTALS-Kyber; FIPS 204 specifies ML-DSA, formerly CRYSTALS-Dilithium; and FIPS 205 specifies SLH-DSA, formerly SPHINCS+.
Jordan Pike AI host
When it released them, NIST urged system administrators to start building the standards into their systems straight away, because full integration takes time. In March 2025 NIST chose HQC, which rests on error-correcting codes rather than lattices, as a backup to ML-KEM, and said it expected a final HQC standard in 2027.
Jordan Pike AI host
NIST warns of harvest now, decrypt later, in which adversaries gather encrypted data today in order to decrypt it once quantum computing matures. And NIST IR 8547, a draft first issued in November 2024, proposes deprecating quantum-vulnerable algorithms such as RSA and elliptic-curve schemes at 112-bit strength after 2030 and disallowing them after 2035.
Jordan Pike AI host
So who should care? Anyone holding data that has to stay secret for a long time. Dex, how does an attacker look at this?
Dex Moreno AI analyst
Harvest now, decrypt later is an attack path with no rush in it. An attacker who could sit on a network path, or who got hold of stored encrypted data, could simply copy it and wait. Nothing needs to break today. The foothold is a pile of ciphertext, and the payoff comes later, when it opens up. The valuable stuff is anything with a long shelf life: customer records, contracts, strategy documents, anything a regulator expects you to protect for years.
Dex Moreno AI analyst
As a technique it's patience as a weapon. Eight out of ten for elegance. The target deserved better, obviously.
Jordan Pike AI host
Dex, we are not rating it out of ten.
Maya Okafor AI analyst
There's no single bug class here, and that's the problem. It's a design assumption that got copied into everything: that RSA and elliptic-curve schemes would stay safe. A partial fix is swapping one product and calling it done. A complete fix starts with an inventory of every place those algorithms live: your own code, the libraries you ship, vendor appliances, and the embedded devices nobody has touched in years. Then cover signatures as well as encryption, because the standards cover both.
Maya Okafor AI analyst
Then check that you can actually change algorithms without a rewrite. If the algorithm choice is hard-coded across your estate, that's the thing to fix first. NIST's own advice is to start now because integration takes time, and that's the part people underestimate.
Dex Moreno AI analyst
Here's where I'd push back. You can't migrate everything at once, so sort by exposure and shelf life. Data that needs secrecy for a long time and travels over networks someone could be recording goes first. Short-lived internal chatter can wait. And IR 8547 is still a draft, so I'd treat its dates as a planning signal, not a firm deadline.
Maya Okafor AI analyst
Agreed on sorting, not on waiting. ML-KEM is a finished standard. HQC is a backup, not a reason to hold off. Any interim control you put in place is a bridge, so give it an end date, or it becomes permanent.
Jordan Pike AI host
So: Monday morning?
Dex Moreno AI analyst
Find the data that has to stay secret longest, and trace where it travels and where it's stored. That's your first migration list, and your board can understand it in terms of which customers and which regulator's clock.
Maya Okafor AI analyst
Start the inventory of RSA and elliptic-curve use, including what your vendors ship, and ask each vendor for their migration plan. Make algorithm choice something you can change by configuration, not by rewrite.





