Many organisations have a post-quantum budget line but no ability to change cryptography quickly, so they risk spending it on one hard-coded migration that will have to be repeated the next time an algorithm, key size or standard changes. Crypto-agility is a capability: the means to find, replace and retire cryptographic algorithms without rewriting systems. Post-quantum cryptography (PQC) is a destination: a specific set of new algorithms, standardised by NIST, that replaces RSA and elliptic curve schemes. This guide compares them as capability and destination, explains why agility is the prerequisite and PQC the first big migration it enables, and sets out a budget order, a worked scenario and the machine identity consequences. It is for security architects, PKI owners and CISOs deciding what to fund in which year. Standards status as checked on 14 November 2026. The design patterns for agile systems are covered in the Crypto-Agility Guide and the migration steps in the Post-Quantum Cryptography Migration Guide; this page covers how the two relate and in what order to invest.
Key takeaways
- Agility is the road, PQC is one journey on it. PQC replaces quantum-vulnerable public key algorithms; agility is what lets you make that change, and the next one, through configuration and policy rather than projects.
- The deadlines are now firm. NIST IR 8547, published as final on 27 October 2026, deprecates 112-bit quantum-vulnerable public key algorithms after 2030 and disallows them after 2035. The NCSC sets 2028, 2031 and 2035 as UK migration milestones.
- PQC without agility is fragile. FIPS 203 already carries a NIST planning note about a future correction, HQC is selected but not yet approved and a FALCON-based standard is still in development. Hard-coding today's choice repeats the RSA-1024 problem.
- Hybrid is an agility mechanism. RFC 10024, published in August 2026, defines hybrid ML-KEM groups for TLS 1.3 whose security holds if either component holds, and lets you drop the classical half later without a protocol change.
- Machine identities carry most of the work. Certificates, SSH keys, code signing keys and workload tokens are where algorithms live in bulk, so certificate automation and short lifetimes are the most valuable agility investment.
The short answer
Crypto-agility, in NIST's definition from CSWP 39, is the set of capabilities needed to replace and adapt cryptographic algorithms in protocols, applications, software, hardware, firmware and infrastructure while preserving security and ongoing operations. (The December 2025 edition was superseded by an update in June 2026; the definition is the anchor either way.) It has no end date. It is measured by how long it takes to change an algorithm across the estate and how much of the estate can be changed by policy.
Post-quantum cryptography is a set of algorithms designed to resist attack by a cryptographically relevant quantum computer: ML-KEM (FIPS 203) for key establishment, ML-DSA (FIPS 204) and SLH-DSA (FIPS 205) for signatures, and the stateful hash-based schemes in SP 800-208. Migration to PQC is a programme with milestones and, eventually, an end.
The relationship is directional. You can migrate to PQC without agility, system by system, by rewriting code and replacing devices, but you will pay that cost again for every future change. You cannot become agile without doing the groundwork PQC also needs: inventory, central policy and automated key and certificate lifecycle. That shared groundwork is why agility comes first in the budget.
Side by side
| Criterion | Crypto-agility | Post-quantum cryptography |
|---|---|---|
| Nature | Capability: engineering and operational | Destination: a family of algorithms and a migration to them |
| Threat addressed | Any algorithm break, deprecation, implementation flaw or policy change | Shor-class quantum attack on RSA, Diffie-Hellman and elliptic curve schemes, including harvest now, decrypt later |
| Defining sources | NIST CSWP 39 and its update | FIPS 203, 204, 205; SP 800-208; NIST IR 8547 timelines; NCSC migration timelines |
| End state | None: maintained continuously | No sole dependence on traditional public key cryptography (the NCSC's definition of full quantum security) |
| Core activities | Inventory, crypto APIs and providers, central policy, automated certificate and key lifecycle, removal of disallowed algorithms | Algorithm selection, hybrid deployment, supplier upgrades, PKI hierarchy changes, device replacement |
| Main cost drivers | Tooling, refactoring to providers, PKI automation, people | Hardware and firmware replacement, larger keys and signatures, interoperability testing, supplier contracts |
| Typical owner | Security architecture and platform engineering | Programme office with PKI, network, application and procurement leads |
| Measure of success | Time to change an algorithm estate-wide; share of systems changeable by policy | Share of priority systems using PQC or hybrid; milestones met |
| Failure if neglected | Every future change becomes a multi-year project | Long-lived data exposed; non-compliant after 2030 and 2035 dates |
Crypto-agility in depth: threat model, architecture and failure modes
Threat model: agility defends against change itself. Algorithms are broken (SHA-1 collisions), parameters age (RSA-1024), implementations fail and standards are revised. The threat is that an organisation learns it must change and cannot do so before the risk lands. A secondary threat is downgrade: supporting several algorithms is only safe if an attacker cannot force the weakest, which ATT&CK tracks as Downgrade Attack (T1689).
Architecture in words: four layers. An inventory that says where each algorithm, key and certificate is used and who owns it. An abstraction layer where applications call a provider, library, KMS or HSM rather than a named algorithm. A policy layer that sets allowed algorithms centrally, for example system-wide crypto policies, load balancer and service mesh profiles, and certificate authority templates. An automation layer that issues, renews and revokes keys and certificates without people. Agility is the sum: when policy changes, the abstraction layer obeys and automation rolls the new keys.
Failure modes:
- Agility in name only: a "crypto library" wrapper that still exposes algorithm names to every caller.
- Negotiation without integrity: many algorithms enabled, old ones never removed, and no downgrade protection.
- Inventory as a one-off spreadsheet: stale within months, never linked to owners or change processes.
- Hardware treated as software: secure boot keys, TPMs and fused trust anchors cannot be changed by policy and need replacement planning instead.
Post-quantum cryptography in depth: threat model, architecture and failure modes
Threat model: a future quantum computer running Shor's algorithm breaks today's public key key establishment and signatures. Confidentiality is at risk now for data with a long secrecy lifetime, because recorded traffic can be decrypted later. Authenticity is at risk later, but signing keys in firmware, PKI roots and devices last for years, so they also need early planning. Symmetric cryptography and hashes are far less affected.
Architecture in words: key establishment moves first, usually as hybrid ML-KEM in TLS, SSH and VPNs, because it protects confidentiality against harvest now, decrypt later and is negotiated per connection. Signatures follow, starting with new trust hierarchies, code signing and firmware, where ML-DSA, SLH-DSA or the stateful hash-based schemes fit different lifetimes and performance needs. PKI, IoT and industrial protocols move most slowly, which the NCSC states directly.
Failure modes:
- Point migrations: one application team hard-codes ML-KEM-768 with its own library, producing a second inventory problem.
- Size surprises: PQC public keys, ciphertexts and signatures are much larger than their elliptic curve equivalents, breaking fixed-length fields, packet sizes and constrained devices.
- Supplier blind spots: products whose cryptography the customer cannot see or change, with no roadmap in the contract.
- Waiting for certainty: delaying discovery until every standard is final, which pushes the hardest work past the 2028 discovery milestone.
Why agility comes first: the sequencing argument
Three facts from the current standards make the case.
- The PQC set is still moving. NIST IR 8547 final lists FIPS 203, 204 and 205 and SP 800-208 as approved, notes that HQC has been selected but is not yet approved, and that the FALCON-based signature standard is under development. FIPS 203 itself carries a November 2025 planning note about a correction in a future revision. An organisation that cannot change algorithms by configuration will face this churn as rework.
- Coexistence is mandatory. The NCSC advises choosing solutions with cryptographic agility because traditional and PQC algorithms will run side by side during migration, and asks organisations to define criteria for when to stop supporting traditional algorithms. That is an agility requirement stated in a PQC document.
- Discovery is the first milestone. The NCSC's 2028 milestone is discovery, assessment and an initial plan; the NIST NCCoE migration project starts with automated discovery and inventory of public key use, then prioritisation using Mosca's theorem. Discovery is also the foundation of agility, so funding it serves both.
The practical consequence: the first euros, pounds or dollars should buy capabilities that make every later migration cheaper. PQC-specific spend that cannot be reused, such as replacing a device fleet, should be scheduled against the dates, not pulled forward ahead of the groundwork.
Budget order: what to fund in which phase
| Phase | Fund | Serves agility | Serves PQC | Exit evidence |
|---|---|---|---|---|
| 1. Discover | Automated cryptographic discovery across network, code, hosts, PKI and cloud KMS; CBOM format; owner mapping | Yes: you cannot change what you cannot find | Yes: NCSC 2028 milestone | Inventory of public key use with owners and data lifetimes for priority services |
| 2. Automate lifecycle | Certificate lifecycle automation, short certificate lifetimes, secrets and key rotation, CA able to issue new key types | Yes: rollout of any new key type becomes routine | Yes: PQC and hybrid certificates can be issued at scale | Share of machine certificates issued and renewed automatically |
| 3. Centralise policy | System-wide crypto policies, TLS profiles at load balancers and mesh, crypto providers or KMS for applications | Yes: one change reaches many systems | Yes: hybrid groups enabled centrally | Share of TLS endpoints governed by central policy |
| 4. Hybrid key establishment | Enable X25519MLKEM768 in TLS where libraries support it; hybrid SSH and VPN as products allow | Partly: proves the negotiation path | Yes: harvest now, decrypt later mitigated for priority traffic | Share of priority external and internal connections using hybrid groups |
| 5. Supplier and procurement | Contract clauses for cryptographic transparency and PQC roadmaps | Yes: future products changeable | Yes: NCSC asks organisations to tell suppliers their needs | Share of critical suppliers with a dated roadmap |
| 6. Signatures and trust anchors | New PQC or composite hierarchies, code and firmware signing, device replacement where keys are fixed | Partly | Yes: authenticity before 2035 | Roots and signing keys migrated per plan |
| 7. Retire | Removal of quantum-vulnerable algorithms from policy and negotiation | Yes: deletion is part of agility | Yes: IR 8547 disallow date | Zero endpoints negotiating disallowed algorithms |
Phases 1 to 3 are the agility core and run first. Phase 4 can start in parallel on systems whose libraries already support hybrid groups, because it is low cost and protects confidentiality now. Phases 5 to 7 follow the dates. Small organisations mostly buy agility through suppliers: keep libraries, operating systems and managed services current, prefer providers that publish PQC roadmaps, and automate certificates. Large organisations need a programme office, a funded discovery tool and an internal crypto platform team.
Hybrid modes as an agility tool
RFC 10024 (Standards Track, August 2026) defines three post-quantum traditional hybrid groups for TLS 1.3. The rationale it gives is that hybrid security holds as long as at least one component algorithm remains secure.
| Group | IANA value | Recommended | Typical use |
|---|---|---|---|
| X25519MLKEM768 | 4588 | Y | General web and service traffic |
| SecP256r1MLKEM768 | 4587 | N | Environments that require NIST curves |
| SecP384r1MLKEM1024 | 4589 | N | Higher security profiles that require P-384 |
Hybrid is agility in miniature: the protocol negotiates a named group, so moving later to pure ML-KEM, or to another KEM, is a policy change rather than a protocol change, provided endpoints are centrally configured. The cost is larger handshakes and a second algorithm to keep patched. Treat hybrid as a transition step with a review date, not a permanent fixture. Protocol detail is in the Hybrid Post-Quantum TLS Key Exchange Guide.
Where identity fits: machine identity impact
Most public key cryptography in an enterprise belongs to non-human identities: TLS server and client certificates, mutual TLS between services, SSH host and user keys, code signing keys, device certificates, SAML and OIDC signing keys, and JWT signing keys for workload tokens. That makes machine identity management the place where agility and PQC meet.
- Short-lived credentials migrate themselves: a workload certificate that lives for hours picks up a new algorithm at the next issuance once the CA and policy change. Long-lived certificates and static SSH keys have to be found and replaced one by one.
- Token signing is a hidden dependency: identity providers and API gateways sign tokens with RSA or ECDSA keys, and every verifier must accept the new algorithm before the issuer switches. Allow-list algorithms per issuer and plan dual-publishing of keys.
- Ownership gaps stall migration: certificates and keys without an owner cannot be scheduled. Fix ownership in phase 1.
See the Post-Quantum Migration for NHIs Guide and the Cryptographic Inventory and CBOM Guide.
Worked scenario: a regional bank
A bank has a board-approved PQC budget and a 2031 target for priority services. The first proposal is to replace VPN concentrators and the payment HSM fleet in year one.
- Reframe: discovery finds 40,000 machine certificates, a third issued manually with two-year lifetimes, and dozens of applications that call RSA directly. Replacing hardware first would leave all of these untouched.
- Year one (agility core): fund automated discovery, CBOM output and owner mapping; extend certificate automation to the manual third and shorten lifetimes; move internal TLS to central profiles at load balancers and the service mesh.
- Year one in parallel (cheap PQC win): enable X25519MLKEM768 on customer-facing endpoints whose libraries support it, protecting long-lived customer data in transit.
- Year two: refactor high-value applications to a crypto provider or KMS; add PQC roadmap clauses to supplier renewals, including the HSM and VPN vendors.
- Years two to four: replace hardware on supplier roadmaps, with the CA already able to issue the new key types and policy able to switch groups centrally.
The same budget, ordered this way, meets the PQC milestones and leaves the bank able to absorb the next standards change, for example HQC approval or a FIPS 203 revision, without another programme.
Detection and measurement
- Inventory coverage: share of priority services with cryptographic use, data lifetime and owner recorded.
- Time to change: elapsed time to change an allowed algorithm or TLS group across a defined scope, measured in an exercise each year.
- Policy-governed share: percentage of TLS endpoints and applications whose algorithms are set centrally rather than locally.
- Automation: percentage of machine certificates issued and renewed automatically, and median certificate lifetime.
- Hybrid adoption: share of priority connections negotiating a hybrid ML-KEM group, from load balancer or network sensor logs.
- Legacy negotiation: count of endpoints still offering disallowed or deprecated algorithms, trending to zero before the IR 8547 dates.
Framework mapping
| Framework | Crypto-agility | Post-quantum cryptography |
|---|---|---|
| NIST CSF 2.0 | ID.AM-02 inventories, ID.AM-08 lifecycle management, PR.PS-02 software maintained and replaced | PR.DS-01 and PR.DS-02 data at rest and in transit, ID.RA-04 impacts and likelihoods, GV.SC-05 supply chain requirements |
| NIST SP 800-53 Rev. 5 | CM-8 System Component Inventory, SA-8 Security and Privacy Engineering Principles, SC-12 Key Establishment and Management | SC-13 Cryptographic Protection, SC-17 PKI Certificates, SA-4 Acquisition Process, RA-3 Risk Assessment |
| ISO/IEC 27001:2022 | Annex A 8.24 Use of cryptography (rules for cryptography and key management) covers both | |
| MITRE ATT&CK v20 | T1689 Downgrade Attack, T1557 Adversary-in-the-Middle (negotiation integrity) | Not technique-specific; threat is future decryption of recorded traffic |
Common failure patterns
- Buying PQC hardware before discovery: large spend on one layer while thousands of certificates and hard-coded calls stay unknown.
- Calling a library upgrade "agility": a new PQC-capable library with algorithm names still hard-coded in application code.
- Hybrid forever: enabling hybrid groups without a criterion for retiring the classical component, which the NCSC asks organisations to define.
- Ignoring verifiers: switching a token or code signing issuer to a new algorithm before every relying party accepts it.
- Citing old dates: planning to the SP 800-57 disallow date of 1 January 2031 for 112-bit schemes, which IR 8547 final has superseded.
- No supplier leverage: renewing contracts for products with fixed cryptography and no dated roadmap.
Practitioner checklist
- Fund automated cryptographic discovery and owner mapping before PQC-specific hardware.
- Record data secrecy lifetimes for priority services to rank harvest now, decrypt later exposure.
- Extend certificate automation to all machine certificates and shorten lifetimes.
- Set allowed algorithms and TLS groups centrally, and make applications call providers or KMS.
- Enable X25519MLKEM768 where libraries support it, with a review date for the classical component.
- Add cryptographic transparency and PQC roadmap clauses to supplier renewals.
- Plan trust anchor and signing key migration separately from key establishment.
- Measure time to change an algorithm each year and report it with PQC milestones.
Standards and references
- NIST: IR 8547, Transition to Post-Quantum Cryptography Standards
- NIST: CSWP 39, Considerations for Achieving Cryptographic Agility
- NIST: FIPS 203, Module-Lattice-Based Key-Encapsulation Mechanism Standard
- IETF: RFC 10024, Post-Quantum Traditional (PQ/T) Hybrid Key Agreement Mechanisms for TLS 1.3
- NCSC: Timelines for migration to post-quantum cryptography
- NIST NCCoE: Migration to Post-Quantum Cryptography
- CISA: Strategy for Migrating to Automated Post-Quantum Cryptography Discovery and Inventory Tools
Related NHI Mgmt Group resources: Crypto-Agility Guide · Post-Quantum Cryptography Migration Guide · QKD vs Post-Quantum Cryptography Guide · NIST IR 8547 PQC Transition Timeline Guide · Crypto-Agility Assessment Template