--- title: "Post-quantum cryptography migration: five steps and the clocks already running | TauX" description: "Governance, CBOM inventory, risk assessment, crypto-agility and supply chain — mapped to the FSC's seven strategies, with the US, EU and Taiwan timelines and where each comes from." url: "https://taux.io/en-US/pqc-migration" locale: "en-US" alternates: ja-JP: "https://taux.io/ja-JP/pqc-migration" ko-KR: "https://taux.io/ko-KR/pqc-migration" zh-Hans-CN: "https://taux.io/zh-Hans-CN/pqc-migration" zh-Hant-TW: "https://taux.io/zh-Hant-TW/pqc-migration" --- # Post-Quantum Migration A five-step path from governance to supply chain, mapped to the FSC's seven strategies, with the US, EU and Taiwan timelines and where each comes from. Contents * [Why now](https://taux.io/en-US/pqc-migration#why-now) * [The five steps](https://taux.io/en-US/pqc-migration#overview) * [00 Governance](https://taux.io/en-US/pqc-migration#governance) * [01 CBOM inventory](https://taux.io/en-US/pqc-migration#cbom) * [02 Risk and compliance](https://taux.io/en-US/pqc-migration#risk-assessment) * [03 Crypto-agility](https://taux.io/en-US/pqc-migration#crypto-agility) * [04 Supply chain](https://taux.io/en-US/pqc-migration#supply-chain) * [Regulatory timelines](https://taux.io/en-US/pqc-migration#regulatory-timeline) * [Standards status](https://taux.io/en-US/pqc-migration#standards) 00 ## Why now Two **independent** reasons, and they are not equally reliable. Keeping them apart matters, because conflating them is how this subject gets oversold. One: harvest now, decrypt later Harvest-now-decrypt-later is a widely accepted threat model: traffic intercepted today can be stored untouched and decrypted whenever the capability arrives. So the question is not whether your data is safe today, but **how long it needs to stay secret**. Anything with a confidentiality horizon beyond ten years — medical records, policies, contracts, identity credentials, trade secrets — is already exposed under that model. One honest caveat: **estimates for when a quantum computer could actually break RSA-2048 vary enormously** — serious people put it anywhere from within a decade to several. Treat any content that states a specific year as settled fact with suspicion. Two: the regulatory deadlines already exist This one **depends on no forecast at all**, and it is the one to plan against. The EU recommends member states begin migration by the end of 2026 and complete high-risk critical-infrastructure scenarios by the end of 2030; the FSC's phased timeline starts in 2026. Those dates hold regardlessof when quantum computers arrive, and they are already written down. Put differently: you do not need an opinion about quantum timelines to work out whether to start this year. 01 ## The five steps Each step is mapped to one of the seven strategies in the FSC's PQC migration guidance for the financial sector. Outside finance there is no local statutory deadline in Taiwan yet, but the sequence holds — and if you sell to a financial institution you are already inside that guidance's scope. | # | Step | FSC strategy | | -- | --------------------------------------------------------------------------------------------- | -------------------------------------------------- | | 00 | [Governance](https://taux.io/en-US/pqc-migration#governance) | ① PQC policy and governance | | 01 | [Inventory the cryptography (CBOM)](https://taux.io/en-US/pqc-migration#cbom) | ② Inventory cryptographic use, build the register | | 02 | [Risk assessment and compliance mapping](https://taux.io/en-US/pqc-migration#risk-assessment) | ⑤ Prioritise migration by risk | | 03 | [Build crypto-agility](https://taux.io/en-US/pqc-migration#crypto-agility) | ③ Improve crypto-agility, remove anti-patterns | | 04 | [Quantum-safe supply chain](https://taux.io/en-US/pqc-migration#supply-chain) | ⑥ Update procurement and supply-chain requirements | The FSC's other two strategies — ④ ecosystem coordination and a shared risk picture, and ⑦ cutover testing and operational resilience — run through all five rather than forming a stage of their own. Step 00 ## Governance What has to be decided Who owns it, where the budget comes from, how often it reports, and which existing committee carries it. PQC migration cuts across security, infrastructure, application development and procurement — no single department can finish it alone. Why this comes before the inventory That is the FSC's ordering, for a practical reason: **a CBOM inventory touches nearly every system and needs cooperation from every team**. Without someone deciding who owns it and where the money comes from, the inventory stalls halfway on "that is not my job". Governance first is what makes the inventory movable — which is why the FSCputs it first in the near-term 2026–2027 phase. In practice 1. 01Hang it off the security governance you already have. A new committee is one more meeting that will not happen. 2. 02Name one owner, responsible for the inventory, the prioritisation and the supplier conversations. 3. 03Budget for it across years rather than as a project. The FSC timeline runs to 2035, which is not an annual budget cycle. Step 01 ## Inventory the cryptography — CBOM — Cryptography Bill of Materials Definition A CBOM is **a register of cryptographic assets**: which algorithms, key lengths, certificates and libraries are in use, where each lives, who maintains it and how long its lifecycle is. It is the cryptographic counterpart of an SBOM,`CycloneDX` has supported the format natively since 1.6. Why everything else waits on it **You cannot prioritise what you cannot enumerate.** Most organisations cannot say how many certificates they hold, which systems still use SHA-1, or which ten-year-old batch job has an RSA key hard-coded in it. Without the register, "migrate the high-risk things first" is not an instruction anyone can follow, because nobody knows which those are. In practice 1. 01Start with external TLS and certificates — easiest to scan automatically, quickest to show a result, and that result is what funds the rest. 2. 02Then cryptographic calls in your own code and its dependencies. This layer is usually the messiest, and where "nobody knows what this does". 3. 03Last, purchased software and hardware. You cannot see inside this layer; you can only ask the supplier — which makes it the same job as step 04. 4. 04Keep it machine-readable and wired into the asset management you already run. **A register that goes stale is more dangerous than none**, because every later decision rests on it. Relation to data governance A CBOM is really the[data governance](https://taux.io/en-US/data-governance) capability — inventory, catalogue, assign an owner, keep it current — applied to cryptographic assets. If you have already done a data inventory, this step goes much faster. Step 02 ## Risk assessment and compliance mapping — Risk-Based Prioritisation The question to answer Once the register exists: what moves first? The FSC's phrasing is "prioritise migration by risk" — not everything at once, which is neither possible nor necessary. Three dimensions to rank on 1. 01**How long the data must stay secret.** Data with a twenty-year confidentiality requirement is already in harvest-now-decrypt-later range today; something that needs three months of secrecy is far lower risk. This dimension is the one most often underweighted. 2. 02**How often the system is replaced.** HSMs, chips and embedded devices can have lifecycles of ten years or more. Buy one this year without requiring readiness and you have locked yourself in past 2035. 3. 03**Compliance exposure.** Against whichever regime applies: financial institutions look at the FSC's phases, cross-border operations at the EU's 2030 critical-infrastructure deadline, and anyone selling into US government at federal procurement requirements. The point of the compliance mapping is not to memorise every date. It is to find **which deadline reaches you first** — and for most companies that is not NIST in 2035, it is a procurement requirement arriving much sooner. Step 03 ## Build crypto-agility Definition Crypto-agility is **being able to change algorithm without changing application logic**. The algorithm is configuration, not something compiled in. Why it outlasts this migration Because **this will not be the last one**. ML-KEM and ML-DSA are the current standards, and NIST is already standardising HQC as a fallback — on the explicit grounds that ML-KEM might one day be broken. Without agility, the next migration repeats the whole inventory and rewrite from scratch; with it, the next migration is a configuration change. The FSC places it in the near-term phase not because it is urgent but because **it sets the cost of every migration after this one**. Anti-patterns to clear first 1. 01Algorithm and key length hard-coded, so changing one means a recompile and a release. 2. 02Home-rolled cryptography instead of calls into a central service or library. 3. 03Certificate and key lifecycles managed by hand with no automatic rotation. If you cannot rotate, you cannot switch. 4. 04Negotiation parameters scattered across dozens of service configs with nothing central. During the transition Most systems will run **hybrid mode** first — classical and post-quantum algorithms side by side — keeping a way back while PQC implementations mature. That is agility paying off directly: being able to run both is what makes the switch safe. Step 04 ## Quantum-safe supply chain Why this one matters especially in Taiwan The FSC guidance applies to **financial institutions and their significant suppliers**. So a company selling software, hardware or services to a bank, insurer or broker is already in scope without being in finance — the pressure simply arrives through a procurement contract rather than a regulator. For Taiwan's tech companies and systems integrators that is a commercial consequence, not a theoretical risk. In practice 1. 01Put PQC readiness into **procurement and tender requirements**. Ten-year hardware signed this year without that requirement is a mine laid for 2035. 2. 02Ask suppliers for **a CBOM or an equivalent account of their cryptography**, rather than accepting "we are looking into it". You cannot see inside packaged software; asking is the only instrument you have. 3. 03Fold supplier timelines into your own prioritisation — your deadline is set by your slowest supplier. 4. 04If you are the supplier: **have the answer ready**. Customers will ask well before the regulatory deadline does. Regulatory timelines ## Where the rules stand **Regulatory status on this page was checked on 2026-08-15.** Every row names the document it comes from. No automated check can hold this section — regulations do not change inside our version control — so this page's last-updated date means **the day the regulatory status was re-checked**, not the day the layout moved. | Jurisdiction | Key dates | Source | | ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | United States | 112-bit-strength algorithms deprecated after 2030, disallowed after 2035 | NIST IR 8547, **still a public draft**. Both years are widely quoted as settled; they are not yet final | | European Union | Member states to begin by end 2026; high-risk critical-infrastructure scenarios by end 2030 by then | Commission Recommendation (EU) 2024/1101, issued 2024-04-11 | | Taiwan financial sector | Near term 2026–2027: governance, inventory method, CBOM, agility foundations Medium 2027–2029: pilots, infrastructure upgrades, shared testing Long term to 2035: migrate high-risk, high-criticality systems first | FSC PQC Migration Reference Guidance for the Financial Sector, issued 2026-06-18. Applies to financial institutions **and significant suppliers**; it is **guidance, not a mandate** | | Taiwan all sectors | **No completion deadline set** | PQC Migration Guidance from the Administration for Digital Industries (moda) and the post-quantum security industry alliance, issued 2025-04-16. Covers PKI, secure chips, digital signatures, network equipment and zero-trust identity | **For most companies here, the thing that reaches you first is not any of these dates. It is a procurement contract.** Financial institutions will start writing PQC readiness into tenders long before 2035, and that is the point at which a supplier actually needs an answer. Standards status ## What is standardised and what is not Separated out because finalised standards and algorithms still in the standardisation process get conflated constantly — and procurement and architecture decisions should rest on the first kind. Final (August 2024) 1. 203**ML-KEM** (formerly Kyber) — key encapsulation, for establishing a shared key 2. 204**ML-DSA** (formerly Dilithium) — digital signatures, the general-purpose first choice 3. 205**SLH-DSA** (formerly SPHINCS+) — hash-based digital signatures,the most conservative security assumptions Still in process **HQC** was selected by NIST on 2025-03-11 as a fifth algorithm, positioned as a **fallback** for ML-KEM should the lattice assumptions ever be broken. **FN-DSA** (formerly Falcon) is likewise still in process. Neither is **a finalised standard you can cite**, so architecture and procurement decisions should rest on FIPS 203/204/205 for now. The existence of HQC is itself the best argument for [crypto-agility](https://taux.io/en-US/pqc-migration#crypto-agility): even NIST is preparing a way out in case today's standard turns out not to be enough. Sources Regulatory timelines and standards status are taken from [NIST Post-Quantum Cryptography Project](https://csrc.nist.gov/projects/post-quantum-cryptography), [Commission Recommendation (EU) 2024/1101](https://eur-lex.europa.eu/eli/reco/2024/1101/oj/eng), [the Financial Supervisory Commission](https://www.fsc.gov.tw/ch/index.jsp) and [the Administration for Digital Industries](https://moda.gov.tw/ADI/news/latest-news/16090) . The five-step mapping follows the seven strategies in the FSC's guidance. The synthesis, terminology and practical recommendations are TauX's and do not represent the position of any of these bodies. ## Not sure what cryptography you are running? Most organisations stall at step one. A CBOM inventory is the same capability as data governance — start by telling us what your asset catalogue looks like today. [Book a consultation](mailto:hello@taux.io) [See data governance](https://taux.io/en-US/data-governance)