An algorithm has to go
Deprecated, broken, or refused by a customer questionnaire. The incident has a clock, and your honest answer is a release calendar across every application that touches it.
ANKASecure© turns the next cryptographic change into a policy operation. Applications invoke stable capabilities against stable key identifiers; three policies govern the mechanism behind them: algorithm availability (list, family, category, security level, standards body, operations), algorithm rotation (no downgrade, no return from post-quantum to classical) and key lifecycle (superseded and retired material keeps decrypting and verifying while new encryption and signing are denied; a key revoked for an algorithm deprecation keeps its history readable, one revoked for compromise is sealed). Re-encryption and re-signing move protected data to a new mechanism without the applications that created it, and every operation is recorded with its algorithm and key. The Cryptographic Response Gap, from the change in a mechanism’s acceptability to the end of its remediation, is set by that architecture rather than by the algorithm.
Nobody asks for crypto-agility by name. What they live with is this: when an algorithm has to change it is a code change, a regression cycle, a security validation and a release — multiplied by every application that uses it. So the change is deferred, and the estate keeps running on what it should already have left behind. ANKASecure© turns that change into a policy operation.
The question is not whether cryptography will change. It is how much of your enterprise has to change when it does.
Is this you
Deprecated, broken, or refused by a customer questionnaire. The incident has a clock, and your honest answer is a release calendar across every application that touches it.
A subsidiary under BSI needs hybrid. A federal contract names CNSA 2.0. An operation in Japan runs on CRYPTREC. Today that means separate stacks, or exceptions nobody wants to sign.
No vulnerability, no incident — just a posture you would rather have. Without a forcing event the change never reaches a sprint, and the gap between what you run and what you would choose keeps widening.
Who signs this: A CISO or head of cryptography who will be asked how fast the organization can move, and an enterprise architect who knows what a change across two hundred applications actually costs. If the answer to “how long to leave this algorithm?” is measured in quarters, this is the conversation.
Why existing approaches fall short
An organization can complete a full post-quantum migration and still not be crypto-agile, because changing the algorithm does not change the architecture that made the change expensive. Key rotation, algorithm migration and crypto-agility are three different capabilities.
What changes
Example
What the application must do
How often it can happen
An organization migrates 200 applications from RSA and elliptic curves to Algorithm X: each one analyzed, its code changed, a new library tested, certified and deployed. Three years, and it succeeds. The year after, Algorithm X stops being acceptable for one of its uses, and the same 200 applications have to be modified again. It completed a migration; it did not achieve agility. Migration answers which algorithm comes next; agility answers how expensive it will be to change that answer again.
Five warnings in four years. Not one required a quantum computer.
Rainbow
SIKE
HAWK
Classic McEliece
RSA signing oracle
The estimated cost of the best published structural attacks on Classic McEliece, for the parameters BSI recommends, as BSI summarized it in September 2026: above 2³⁰⁰⁰ for decades, and 2⁹⁹ to 2¹⁰² in a 2026 preprint. The newest results have not completed peer review and none enables a practical attack, yet BSI advised against Classic McEliece in new developments. Three of the five cases were candidates that never became standards, and ML-KEM and ML-DSA are not affected by any of them. The point is that the assessment moves, on its own schedule.
How it works
Applications express what they need; policy decides which mechanism serves it, and enforces that decision where the cryptography executes.
Encrypt, decrypt, sign and verify against a stable key identifier. A rotation advances to a new material version under the same identifier, and that version may use a different algorithm within the same purpose family, with no change to the application.
Which algorithms may be used: by explicit list, family, category (classical, post-quantum or hybrid), security level, mandatory standards body or permitted operations. A hybrid-only profile implements a BSI-style requirement as policy rather than code, and a change propagates in under a second.
Which transitions are allowed, for example forbidding a security downgrade or a return from post-quantum to classical algorithms. The server computes whether a move is admissible before you make it.
Which operations each state permits. Superseded and retired material keeps decrypting and verifying while new encryption and signing are denied; a key revoked because its algorithm was deprecated keeps its history readable, while a key revoked for compromise is sealed on every operation.
Information protected under an old mechanism moves to a new one without the applications that created it, including streaming operations for large objects. The exit becomes a process, not a program.
Every operation is recorded with its algorithm and key in a signed audit trail. Showing that a mechanism has applied no new protection since a given date is a query, not an audit of individual applications.
Agility done right
Adam Langley argued that extensibility adds code, bugs and dangerous interactions, and tends to fail when it is finally needed. RFC 7696 warns that negotiation creates downgrade attacks and that supporting too many algorithms is harmful. NIST acknowledges that agility adds complexity. They point to a specific kind of agility, not away from it.
The mechanism for a use is decided by policy and enforced centrally, not negotiated per session between parties that can be pushed toward the weakest option.
A small set of approved mechanisms per use. Agility does not mean supporting everything.
Change happens at a single governed point, used routinely, so that it works when it is needed.
Deprecated mechanisms are removed from new use deliberately, not left lying around as fallbacks.
Hybrid constructions, which combine a post-quantum mechanism with a classical one so that both must be broken, are a hedge: BSI notes that its hybrid recommendation proved its worth when Classic McEliece weakened. Hybrid buys time; agility is what makes hybrid configurations practical to introduce, adjust and retire.
The cryptographic response gap
The Cryptographic Response Gap runs from T0, when the assessment, the standard or the acceptable-use status of a mechanism changes for a given use, to T1, when the organization has finished remediating it: no new protection applied with it, and the information it protected handled according to policy. It specializes Schneier’s window of exposure and the migration-time variable of Mosca’s inequality.
Two organizations that use exactly the same algorithms face very different risk if one can stop new use of a mechanism across its estate through a governed policy change and the other has to open a multi-year program to do the same.
Detection
Exposure analysis
Decision
Engineering
Deployment
Exit
Discovery and inventory programs shorten detection and exposure analysis, and they are necessary. In a typical enterprise the longest part of the gap is engineering, deployment and exit, and those are set by architecture. That is also where the gap really lives: transport cryptography negotiates its algorithms per session and changes at a few endpoints, while data at rest and in use is protected with long-lived keys and with choices made inside application code.
Twenty-nine templates in the catalogue — NIST_APPROVED, EU_COMPLIANT for ENISA, GDPR, NIS2 and eIDAS 2.0, HIGH_SECURITY for levels 3 and 5 — applied deployment-wide and overridable per tenant. Changing the template changes what the next operation may use. Below it, the catalogue the templates choose from: every algorithm with its category, level, standards and status, Classic McEliece already marked legacy.
Fifteen minutes
Live, on a real deployment, with an algorithm of the kind you are stuck on. Nothing in this list is a mock-up or a roadmap item.
An application encrypting under a classical algorithm, referencing its key by name.
The admissibility verdict for that key: which candidates are allowed, which are refused, and the reason code for each.
The policy changes centrally. No application is touched. No release is cut.
The same application, unchanged and not restarted, now writing under the new mechanism.
A candidate the policy refuses: the operation is denied with its reason, never silently downgraded.
The audit trail: which policy decided which operation, signed — the evidence an auditor asks for.
Broken is not the threshold
Enterprises do not live on the line between secure and broken. They live on acceptability, and acceptability can be withdrawn long before anything breaks. NIST SP 800-131A assigns each algorithm a status per operation: three-key Triple DES can no longer encrypt but can still decrypt, and SHA-1 cannot generate new signatures but can still verify old ones. A controlled exit needs those states, expressed per operation.
Acceptable
Not for new deployments
Deprecated
Legacy use
Migration only
Disallowed
NIST IR 8547, final since October 2026, deprecates quantum-vulnerable public-key cryptography after 2030 and disallows it after 2035, whether or not a quantum computer has been demonstrated by then. BSI’s decision on Classic McEliece took months.
A mechanism in legacy use cannot encrypt or sign even if an application asks it to: the state is enforced at the operation, not by convention.
The organization can prove that a mechanism has applied no new protection since a given date, and that its remaining use is confined to the workflows policy allows.
What it rests on
Shipping
Stable key identifiers
The application references a kid; material versions move underneath it and historical data keeps decrypting under the version that wrote it.
Shipping
Algorithm availability policy, two planes
Deployment-wide or per tenant, enforced on every operation, fail-closed: an unreadable policy denies rather than permits.
Shipping
Twenty-nine policy templates
Jurisdictional and standards-aligned sets over a catalogue of 120+ algorithms, 33 of them post-quantum.
Shipping
Server-computed rotation admissibility
A verdict per candidate with a reason code, produced by the same rule that runs when the rotation executes.
Shipping
Sub-second policy propagation
A change reaches the enforcement point per tenant without a restart or a redeploy.
Shipping
Signed audit trail
Which policy decided which operation, tamper-evident, forwardable to your SIEM with the signature preserved.
The framework
The Cryptographic Control Plane reference architecture describes each pillar through the scenarios it must handle and one practical test. This solution is two of the three scenarios of the first pillar — responding to an algorithm vulnerability, and evolving the cryptographic strategy without an application release.
01
Primary pillar
Scenario realised
Responding to an algorithm vulnerability
Practical test
“Can the cryptographic strategy change without requiring an application change or release?”
04
Supporting pillar
Scenario realised
Enforcing a new enterprise security requirement
Practical test
“Can the organization prove, by querying the control plane, which policy applied to a given operation, and change that policy centrally?”
Maturity move
From a cryptographic change that needs a release to one that is a policy operation — and from there to an organization that can state how long a transition takes, because it has already done one and has the evidence.
It does not apply to applications that have not been onboarded. The reference architecture is explicit about this and so are we: onboarding is real work. What changes is the economics of every transition after it — the first application is a project, the next change is not.
It does not make a representation universally interoperable. An external system can read what you produce only when it supports that operation’s algorithms, identifiers and profiles; agility on your side is not adoption on theirs.
It does not decide for you what is acceptable. The policies enforce the states your standards bodies, regulators and auditors define; tracking those assessments and choosing the replacement remain decisions of the people who own your cryptographic risk.
Where to start
The worst time to think about crypto-agility is after the post-quantum migration. If you are going to open the code of your applications anyway, open it once: replace the embedded algorithm with a stable capability governed by policy, and the migration becomes the last cryptographic change those applications need.
1
Identify the mechanisms and systems that would define your gap if a mechanism you rely on became unacceptable.
2
ANKASecure© changing an algorithm, retiring a mechanism and re-encrypting data end to end, on a case analogous to yours.
3
One critical use governed by policy, with keys anchored in your existing HSM or KMS.
Whitepaper
Five real cases from the last four years, acceptability as a status, the Cryptographic Response Gap, the 200-application problem, how to exit old cryptography, and six questions for the CISO. 27 pages, November 2026.
Read it online →Download the PDF
The download opens on this page as soon as the form is submitted. We ask for a corporate account so we can route follow-up to the right team.
After this
No new deployment and no second contract for the platform. Once cryptography is a policy operation, the rest is a change of policy.
Bring the one you would most like to be rid of. Fifteen minutes on a live deployment: the admissibility verdict, the policy change, and the same application writing under the new mechanism without a release.
Book the demo