Seven out of ten Swiss financial institutions have done nothing about quantum risk. No measure planned, none implemented. That is FINMA's own number, from its own survey, and in July it published it next to a date: mid-2027.
Much of the commentary on Guidance 05/2026 since then gets the status wrong in one of two directions. Some call it a new mandatory regime. Others note that it is only guidance and file it under next year. The second mistake is the expensive one.
FINMA did not write a new rule. It said that the rules you already have on governance, operational risk and resilience are technology neutral, and that they already cover a quantum computer able to break your encryption. Then it added two things: numerous institutions need to act to keep meeting those rules, and the topic will get more attention in ongoing supervision. Translated from regulator into plain English, you are already on the hook, and they are going to ask.
They believe it and they are waiting anyway
FINMA surveyed 60 banks, insurers, managers of collective assets and financial market infrastructures between November 2025 and January 2026. Around two-thirds expect quantum cyber risk to affect them directly within seven years. Around two-thirds also think that within ten years a quantum computer will crack RSA-2048 in under 24 hours.
And yet 72% had not planned or implemented a single measure. 28% had a strategic decision at executive or board level, 20% had a project running, and 8% had a roadmap. 43% had not even decided whether to write one.
That gap says more than the deadline does. These are not institutions that doubt the threat. They have put a date on it themselves. Security teams know the pattern, and it always means the same thing: the risk has no owner. Everybody agrees it is serious and nobody has it in their objectives.
So, FINMA is assigning the owner. The guidance asks for five things:
- A strategy adopted by the board of directors, with milestones and target dates for critical business processes and for full migration, and a PQC roadmap by mid-2027 at the latest.
- A risk analysis of every business process and a cryptographic inventory that is kept up to date.
- Priority for data that must stay confidential, intact, or provable for a long time.
- Crypto-agility in every system you buy or build.
- The same from your service providers, because the responsibility for what you outsource stays with you.
Note what mid-2027 is. It is the date for the roadmap. FINMA sets no date for finishing the migration. Each institution sets that one itself, in writing, with its board's signature on it, and then gets measured against it. That is arguably harder than being handed a deadline.
Quantum key distribution and the risks of using quantum applications are out of scope. This is about replacing algorithms.
The algorithms on the list
FINMA names RSA, ECDSA, EdDSA, DH and EC-DH as examples of quantum-vulnerable algorithms. If that reads like the table of contents of your infrastructure, it is.
All five are public-key cryptography, the part that negotiates keys and signs things. Symmetric encryption and hashing get weaker against a quantum computer and you compensate with longer keys. Public-key schemes get broken, and the only fix is to replace them. The replacements exist. The guidance points to NIST's FIPS 203, 204, and 205.
So do not start with your encrypted databases. Start with TLS, VPNs, certificates, code signing, HSM key wrapping, authentication tokens, and every signature you might one day have to show a court.
There are two clocks running and they are easy to confuse. Confidentiality is already exposed: someone records your encrypted traffic today and opens it when the machine exists. Harvest now, decrypt later. Nothing you deploy in 2032 protects what left the building in 2026, which is why FINMA wants long-lived data moved first. Signatures run on the other clock. Nobody can forge one until the machine exists, and that buys time, except for the signatures that need to be trusted for decades.
The arithmetic fits on a napkin. Take the years the data has to stay secret and add the years the migration will take. If the total is longer than the time until a capable quantum computer, the institution is late already. The 8% with a roadmap usually foresee four to five years until critical data and processes are quantum-safe. Now think about a client relationship that is expected to stay confidential for thirty.
A word on hybrid schemes, because the point is often misquoted. FINMA does not mandate them. It notes that various organizations recommend pairing a classical algorithm with a PQC one in the short to medium term, says this adds protection and also complexity and implementation risk, and leaves the call to each institution's own risk analysis. Halborn's position: hybrid makes sense for key exchange on anything long-lived and confidential, provided the complexity is budgeted for. Two algorithms are two things to get wrong.
Crypto-agility, or where migrations go to die
FINMA defines crypto-agility as the ability of a system or application to swap cryptographic algorithms flexibly, without major changes to the software architecture. It recommends making that a requirement for everything you procure or develop.
This is probably the line in the guidance that will age best. Migrating once is a project with an end date. Being able to do it again is a property of the architecture, and it will be needed. The new algorithms are young. Nobody serious will promise they survive the next twenty years. An institution that hardcodes ML-KEM where RSA used to be has spent the budget and kept the problem.
Where does agility die? Almost never in the crypto library.
It dies in the HSM, which runs the algorithms its firmware supports and nothing else. It dies in sizes: PQC keys and signatures are several times larger than elliptic curve ones, and your message formats, certificate fields and database columns were dimensioned for the small ones. It dies in protocols, because threshold signing and MPC schemes are designed around one signature algorithm, and changing it means changing the protocol and auditing it again. And it dies in the org chart. Applications pick the algorithms, infrastructure runs the PKI, a vendor holds the keys, and nobody owns the whole thing.
A policy that says systems shall be crypto-agile is worth the paper. Nobody finds out whether a procedure works by reading it. The test is a drill: pick one production system, swap one algorithm inside a normal release cycle, and write down how long it took and what broke. That write-up is worth more in front of a supervisor than any policy.
Then, there are your providers. FINMA recalls that under Circular 2018/3 the responsibility for outsourced functions stays with the outsourcing institution. It wants crypto-agility as a prerequisite in new software and data outsourcing arrangements, and added to existing ones at the earliest opportunity. If your core banking, your cloud and your custody technology all come from someone else, most of your roadmap is a list of other companies' roadmaps. Get theirs in writing, with dates. A slide that says quantum-ready is not a roadmap.
The paragraph on digital assets that nobody quoted
Most of the commentary skipped one clause. For anyone in digital assets, it is the one that matters.
The scope of FINMA's risk analysis expressly includes distributed ledger technology, with a reference to Article 973d of the Code of Obligations, the provision behind ledger-based securities. It is one clause in one sentence. For an institution that holds digital assets in custody or issues tokenized instruments, it puts that business inside the quantum roadmap. What follows is Halborn's reading of what that means in practice. FINMA does not spell it out.
Everywhere else in a bank, a vulnerable algorithm protects something. In digital assets the algorithm is the asset: whoever can produce the signature owns the funds. The major networks sign with ECDSA and EdDSA, two of the five names on FINMA's list. And you did not choose them. The network did.
For a custody team, the problem splits in two.
The first half is what you control: key generation, the HSM and MPC infrastructure, the encryption that wraps key shares and backups, the channels between signers, the approval layer. That migrates on your schedule and belongs in your roadmap with dates.
The second half is what you depend on: the signature scheme of each network and the upgrade path of each token contract. That migrates on someone else's schedule. Your job is to track it, push for it, and know what you will do if it does not come.
Two risks in the first half almost never make it into a risk register. One is backups. An encrypted backup of key material is the perfect harvest now, decrypt later target: a backup wrapped with RSA and copied today is a private key delivered in a few years. Backups exist to reduce the risk of losing keys, and every copy widens the risk of someone else getting them. Quantum makes that trade worse.
The other is public key exposure. A ledger is public and permanent, so there is nothing left to harvest. Every address that has revealed its public key on-chain is a known target on the day a capable machine exists. Address hygiene, limits on key reuse and the ability to sweep funds to fresh keys fast are controls you can put in place this quarter.
For tokenized instruments, upgradeability is your agility. Who holds the admin keys, under which algorithm, and whether the contract can adopt a new signature scheme all belong in the inventory. So does the uncomfortable answer when the contract cannot.
What to have on the table by mid-2027
Order matters here. A roadmap written before the inventory is fiction, and boards can smell it. Do the inventory, let it drive the risk analysis, and only then go to the board. Each piece needs a name and a piece of evidence, because those are the two things a supervisor asks for.
| What | Who owns it | What proves it |
|---|---|---|
| Cryptographic inventory covering in-house, outsourced, cloud and DLT systems | CISO with IT architecture | A living register by business process: algorithm, key length, system owner, provider, and how often it is refreshed |
| Data ranked by how long it must stay protected | Data owners with risk management | The list of long-lived data sets and the systems that carry them |
| Board-adopted strategy and PQC roadmap | Board of directors, sponsored by the CISO or CRO | Board minutes and a roadmap with milestones and target dates for critical processes and full migration |
| Crypto-agility as a requirement | CTO with procurement | The clause in RFPs and architecture standards, plus one algorithm swap rehearsed in production |
| Providers on the same path | Outsourcing management with legal | Dated PQC roadmaps from each critical provider and the clause in new contracts |
If you run a digital asset business, add a sixth line owned by whoever runs custody: signature scheme, public key exposure and upgrade path for every network and every contract you touch.
One caveat. How this guidance weighs on a specific license and supervisory category is a legal reading, and Swiss counsel or the institution's FINMA contact should confirm it. The direction is not in doubt. FINMA has said in writing that the topic will get more prominence in supervision, and only 8% of the sector had a roadmap when it asked.
Mid-2027 is close. The inventory alone takes months in most institutions, and longer once providers are involved. If there is time for only one thing this quarter, start it. For institutions whose exposure runs through custody or tokenization, this is the work Halborn does every week.
