Your 2035 Problem: Quantum Risk in Validated Systems
Your electronic signatures have a retirement date. Nobody sent you the memo.
By Tushar Arora · November 20, 2026
Written by a pharma QA and compliance practitioner with 14+ years in the industry — for QA, CSV, and compliance leaders at small and mid-size pharma companies, CDMOs, and biotechs. This article is an abridged version of our free flagship report of the same name.
The short version
- The locks your validated systems rely on — RSA and elliptic-curve cryptography (ECC) — are on a published retirement timetable. Under NIST's draft transition plan (IR 8547), they are to be deprecated after 2030 at today's common security levels and disallowed after 2035.
- Pharma's problem is worse than most industries': your electronic signatures and records must stay trustworthy for decades, and changing cryptography inside a validated system triggers revalidation.
- The first step is not a maths project. It is a list — an inventory of every place your systems use this cryptography. Almost nobody has one: in October 2026, the GAO reported that only 1 of 24 major federal agencies had fully addressed all the activities for developing a prioritized inventory of its vulnerable cryptography.
- The FDA has not mandated post-quantum cryptography for GxP systems. This article will not pretend otherwise. The case for starting now rests on arithmetic, not fear.
1. The lock story
Encryption is a lock on data. When an analyst signs a batch record electronically, when a laboratory system stores results, when a manufacturing system sends data between sites — a mathematical lock scrambles the information so that only the right key can read it, or proves it has not been changed since it was sealed.
Today's two common lock designs are called RSA and ECC (elliptic-curve cryptography). Nearly every system you run uses one or both of them, usually without saying so on the label. They sit underneath electronic signatures, underneath the secure connections between systems, underneath the encryption that protects databases and backups.
A quantum computer is a new kind of machine that works on different physics from a normal computer. For most jobs it is not better. For one specific job — the maths problem RSA and ECC are built on — a powerful enough quantum computer will be dramatically better. The method that does it has a name, Shor's algorithm, and it has been known since 1994. What has been missing is a machine powerful enough to run it at scale. That machine does not exist yet. Nobody honest can tell you the year it will.
So why act now? Because of a practice security people call "harvest now, decrypt later." Attackers are already copying encrypted data — intercepted traffic, stolen backups, exfiltrated archives — and storing it. They cannot open it today. They are betting they will be able to open it the day a capable quantum computer exists. For data that only needs to stay secret for a year, that bet matters little. For data that must stay trustworthy for twenty or thirty years — batch records, clinical data, electronic signatures that must still prove who approved what, decades from now — the risk does not start when the machine arrives. It started the day the data was copied.
Picture it as a house. Every lock becomes pickable on some future date nobody can circle on a calendar. Burglars are already walking the street photographing your keys. Changing every lock will take years, because in your house — a regulated house — every lock change needs an impact assessment, testing, and a signature. The sensible first move is not to panic, and not to buy new locks at random. It is to walk the house and list every lock you own.
That list is the inventory. Everything else in this article follows from it.
2. The deadline is real, and it has dates
You do not need to believe a quantum computer arrives on any particular date. The retirement dates for the old locks are already published, and institutions are moving to them.
The replacements exist. In August 2024, NIST published the first three finalized post-quantum cryptography standards: FIPS 203 (ML-KEM, for key establishment), FIPS 204 (ML-DSA, for digital signatures), and FIPS 205 (SLH-DSA, a hash-based digital signature standard). Post-quantum cryptography (PQC) is not experimental anymore; it is standardized. The work now is finding and replacing the old locks.
The timetable is published. NIST's draft transition plan, IR 8547 (initial public draft, November 2024), sets the timetable the U.S. federal migration is now following: quantum-vulnerable algorithms such as RSA and elliptic-curve cryptography at the 112-bit security level — the level most systems use today — are to be deprecated after 2030, and all of these algorithms are to be disallowed after 2035. Two honest qualifications. First, IR 8547 is a draft transition plan, not a law, and it binds federal practice rather than your company directly. Second, "deprecated" does not mean "stops working on January 1, 2031" — it means the algorithm may still be used but carries acknowledged risk; "disallowed" means it is no longer permitted for the stated purpose, with legacy use such as verifying existing signatures treated separately. The direction, however, is not in doubt, and the dates are the anchor everyone is planning against.
The federal machine is already moving. OMB Memorandum M-26-15, "Execution of the Migration to Post-Quantum Cryptography" (June 24, 2026), requires U.S. federal agencies — national security systems excluded — to submit a post-quantum cryptography migration plan to OMB and the Office of the National Cyber Director no later than 120 days from the date of the memorandum: October 22, 2026. Those plans run in phases through 2035, aligned to the NIST timetable.
Why should a pharma QA director care about a federal paperwork deadline? Because federal requirements flow downhill — as supplier questionnaires, sponsor requirements, and customer audits. If your company supplies a federal customer or answers due-diligence questionnaires from larger sponsors, the question "what is your post-quantum cryptography plan?" is coming through that channel: not as a regulation, but as a form you will have to fill in. Cyber insurers are beginning to ask about cryptographic readiness as well. CyberCube's Cyber Predictions 2026 notes that insurers are starting to explore cryptographic-readiness indicators — focused for now on education and scenario analysis rather than pricing — which means the question is likely to reach policyholders before it reaches them as a price.
Your vendors, customers, and insurers will work to the published dates whether you do or not. The only choice is whether your plan exists before the questionnaire arrives, or after.
3. How long is "safe"? The arithmetic nobody can argue with
The question every leadership team asks is: how long are we safe? It deserves an honest answer: nobody can date the arrival of a cryptographically relevant quantum computer, and anyone who gives you a year is selling something. We do not sell predictions. We use arithmetic instead.
A researcher named Michele Mosca reduced the problem to one inequality. Give three numbers:
- X — how many years your data must stay trustworthy.
- Y — how many years the migration will take, from starting the inventory to finishing the replacement.
- Z — how many years until a quantum computer can break today's cryptography.
If X + Y is greater than Z, you are already late — not when the machine arrives, but now, because data harvested today must survive until the machine arrives and your migration has to finish before then.
Run it for a typical GxP record. A batch record or signature that must remain verifiable for the life of the product and beyond — say 25 years — gives X = 25. A realistic migration for a mid-size pharma company, including vendor timelines and revalidation, is no weekend project: Y of 5 years is optimistic. So X + Y is 30. To be safe, Z would have to be more than 30 years away — a claim nobody can make. Notice you never needed to predict Z at all. The inequality fails under almost any plausible assumption, because X in pharma is enormous. That is the whole argument, and it does not depend on hype.
This is also why "we will just use bigger keys" is not a plan. Shor's algorithm defeats the underlying maths at every key size — a bigger RSA key buys time against ordinary computers, but against the machine this article is about, it buys essentially nothing. The only fix is a different lock design: the standardized post-quantum algorithms.
4. The first step is a list, not a maths project
If the algorithms are standardized and the dates are published, what is actually stopping companies? Not mathematics. The bottleneck is embarrassingly basic: most organizations do not know where their vulnerable cryptography is.
The evidence comes from the most documented IT estate on earth. In October 2026, the U.S. Government Accountability Office (GAO) reported that only 1 of the 24 major federal (CFO Act) agencies had fully addressed all of the activities for developing a prioritized inventory of its systems with vulnerable cryptography — and none had fully identified the funding needed to transition its priority systems to post-quantum cryptography (GAO-27-108740). These are agencies with architecture offices, compliance staff, and years of warning. The other twenty-three had addressed the inventory activity only partially or not at all. Partially is where almost everyone lives: fragments of the picture, in different formats, owned by different people, none of it joined up.
"But we have architecture diagrams"
An architecture blueprint is drawn at system level: boxes for systems, arrows for interfaces, notes on data flows. Cryptography lives below blueprint level. The blueprint says "the LIMS connects to the historian over a secure interface." It does not say which algorithm that interface negotiates, at what key length, with which certificate, issued by whom, expiring when. It does not say which algorithm seals the electronic signatures inside the application, because that choice is buried in the vendor's product, not drawn on anyone's diagram. And blueprints are drawn at design time; systems drift.
A blueprint is a map of the house. The inventory is the list of locks — room by room, lock by lock, with the make and model of each. Different documents, different jobs.
A five-question self-test
- Which of your systems sign records electronically?
- For each one, which algorithm and key length does the signature use?
- Which of your vendors has given you their post-quantum roadmap in writing?
- For each system's records, how long must the signatures remain verifiable — and who decided that number?
- If you changed the cryptography inside one validated system tomorrow, what exactly would break, procedurally?
If you scored five out of five, you are ahead of 23 of 24 federal agencies. Most experienced QA leaders we describe this to score zero to two. That gap — not the quantum computer — is the immediate problem, and it is entirely fixable with ordinary governance work.
5. Why pharma cannot copy what the banks are doing
First, Part 11 longevity. Under 21 CFR Part 11, electronic records and signatures must be trustworthy and retrievable for as long as they are required. To be precise — and precision matters here — the retention periods themselves come from predicate rules and your company's own SOPs, not from Part 11. What Part 11 sets is the trustworthiness requirement: the signature must remain linked to its record, attributable, and verifiable for the whole of that retention life. In practice that life is long — a signature applied in 2026 may need to be provable in 2051. This is X in the Mosca arithmetic, and in pharma X is measured in decades.
Second, the revalidation trap. In a validated system, the cryptography is not an IT detail — it is part of the validated state. Change the algorithm that signs records or encrypts the database and you have disturbed that state: impact assessment, risk assessment, testing, documentation updates, QA review and approval. IT cannot "just upgrade the encryption over the weekend," and any migration plan that assumes it can will fail at the first change-control board. This is why Y in the arithmetic — the migration duration — is longer in pharma than anywhere else.
Third, the historical records question. Migration plans talk about the future. GxP lives in the past as well: millions of records and signatures created under RSA and ECC, which must still verify after you migrate. Will they? Under what mechanism — re-signing, timestamping, vendor tooling, or legal reliance on legacy verification? Almost nobody has asked their vendor this question in writing.
Fourth, the CSV documentation gap. No standard CSV deliverable records cryptography at per-use level. What exists are fragments: a design specification may state "TLS 1.2" or "AES-256" for a system as a whole — one line, at system level, written at go-live; a Part 11 or Annex 11 assessment checks signature controls, not which algorithm or key length performs the signing; vendor assessments ask encryption questions as yes/no checkboxes. The joined-up register — per use: this system, this algorithm, this purpose, this retention horizon, this validated-state consequence — is exactly what a crypto inventory is.
Generic vendors can replace your algorithms. They cannot tell your QA director what that does to your validated state. That sentence is the reason this practice exists.
6. A sensible order of work: Find → Judge → Plan
Find. Build the inventory — one row per use of cryptography, not per system. A LIMS that signs records, encrypts its database, and connects over TLS is three rows, because the three uses carry different risks and migrate in different ways. Every row records the system, algorithm, purpose, retention horizon, Part 11 relevance, validated-state impact, vendor PQC status, and where the information came from.
Judge. Score each row on three factors, 1 to 3 each: data longevity (how long must this stay trustworthy?), exposure (how reachable is it?), and validated-state impact (what does changing it trigger?). The scores add to a band — high, medium, low — and the band sets the migration order. Any item with direct Part 11 relevance gets called out by name in the findings regardless of its total, and any factor that cannot be established is scored conservatively — an unknown never gets the benefit of the doubt.
Plan. Sequence the migrations against the published timetable: high-priority items planned and scheduled well before the after-2030 deprecation; everything resolved before the after-2035 disallowance. Ride planned upgrades wherever the change calendar allows. Items blocked on a vendor become vendor-watch items with a named next check date, not items quietly dropped from the plan.
7. What to do Monday morning
Five things, none of which requires buying anything:
- List your GxP systems and their owners. E-signature platforms, LIMS, MES, QMS, and the systems that feed them. If this list is hard to produce, that is your first finding.
- Send the vendor questions to every GxP vendor, in writing. The full report (Section 7) has the exact five questions. Log who answers, who deflects, and who goes quiet.
- Mark which systems sign records under Part 11. Those are your longest-X items — the ones the Mosca arithmetic hits first.
- Write down your retention horizons per system, from your SOPs and predicate rules. You cannot score longevity without them.
- Name one owner for "PQC readiness" inside your company — before a customer, sponsor, or insurer names the deadline for you.
Do those five and you will be ahead of most of the industry — including, on the GAO's evidence, most of the federal government.
What we do not know — and will not pretend to. We do not know when a cryptographically relevant quantum computer will arrive. We do not know whether NIST's timetable will move, or whether the FDA will speak on post-quantum cryptography for GxP systems — as of this article, it has not. We are not cryptographers, and Quantum-Ready GxP does not design algorithms or perform hands-on remediation; implementation is partnered out or referred to specialists. What we do know: the replacement standards are published, the retirement timetable is published, the federal migration is underway, and the inventory — the agreed first step — is unfinished almost everywhere it has been audited, including in 23 of 24 major federal agencies.
Read the full report — free
This article is the abridged version. The full report, Your 2035 Problem: Quantum Risk in Validated Systems, adds the discovery tiers (what scanning can and cannot find), the five vendor questions in full, and the honest cost-of-waiting analysis. Enter your email and the download starts right away.
We store your address to send occasional PQC-readiness notes for pharma QA leaders. No spam, no list resale; unsubscribe any time.
Start your own inventory
The Template Pack gives your QA team the GxP crypto-inventory workbook plus the vendor PQC questionnaire — the starter tools to begin your own inventory and ask your software vendors the right questions.
Sources: NIST IR 8547, Transition to Post-Quantum Cryptography Standards, initial public draft, November 2024 (deprecated after 2030 at 112-bit strength; disallowed after 2035); NIST FIPS 203/204/205 (August 13, 2024); GAO-27-108740 (October 6, 2026); OMB Memorandum M-26-15, "Execution of the Migration to Post-Quantum Cryptography" (June 24, 2026); CyberCube, Cyber Predictions 2026 (January 2026). © 2026 Quantum-Ready GxP · quantumreadygxp.com