The Key That Breaks Everything

The Architecture of Trust

The Invisible Foundation of Global Finance

Every time money moves electronically - every wire transfer, every payment card authorisation, every bond settlement, every inter-bank communication - it passes through a door that is locked with mathematics. Not with a password or a secret key that someone memorised, but with a problem so computationally hard that solving it, even with every computer on earth working in parallel, would take longer than the age of the universe. This is not a figure of speech. It is, or has been until very recently, a literal statement of fact.

The two mathematical problems underpinning essentially all of modern financial cryptography are integer factorisation and the elliptic curve discrete logarithm problem. The first is the basis of RSA (Rivest-Shamir-Adleman), the algorithm that has encrypted internet traffic since 1977. The second is the basis of ECC (Elliptic Curve Cryptography), which powers digital signatures, cryptocurrency wallets, and the authentication systems that let banks prove they are actually talking to each other and not to an impersonator.

ECC works on a different kind of hard problem: given a point on an elliptic curve and a destination point, find the scalar that maps one to the other. It is similarly intractable by classical means. And crucially, ECC underpins ECDSA, the signature algorithm that every Bitcoin and Ethereum transaction uses to prove that you - and only you - authorised a transfer from your wallet. It also underpins TLS 1.3, which encrypts the session between your browser and your bank.

The Bank for International Settlements estimated global foreign exchange turnover at $7.5 trillion per day as of 2022. Add inter-bank settlements, bond markets, derivatives, and payment networks, and you are describing a number so large it defies intuition. Every dollar of it travels through doors locked by these two mathematical problems. The problems have held for decades. They are now facing a different kind of solver.

The Quantum Shortcut

Shor's Algorithm and the End of the Hard Problem

In 1994, mathematician Peter Shor published an algorithm that, given a sufficiently powerful quantum computer, could factor large integers in polynomial time. This is not a modest improvement on classical factoring methods. It is, theoretically, a different category of machine doing a different category of computation. The hard problem that RSA depends upon - the one requiring 300 trillion years on a classical computer - becomes tractable on a quantum computer running Shor's algorithm in hours or days.

A quantum computer exploits superposition, holding many possible states simultaneously, and entanglement, linking the states of multiple quantum bits (qubits) so that operations on one instantly affect others. The result is that certain problems which require a classical computer to try billions of paths sequentially can be explored in massive parallel by a quantum machine. Integer factorisation is one of those problems. The elliptic curve discrete logarithm is another.

The catch, for most of the last thirty years, was the qubit count. Shor's algorithm needs not just qubits but reliable, error-corrected qubits - and the estimates for how many you would need to break RSA-2048 ran into the tens of millions. That number made the threat theoretical and far-off. In 2024, those estimates still stood at roughly 20 million physical qubits. Then the research began to move quickly.

Three significant research papers published between January and March 2026 have rewritten the timeline estimates. The quantum resources needed to break RSA have dropped by an order of magnitude. Some newer architectural approaches put the threshold below 100,000 qubits - a number that existing and near-term quantum hardware is beginning to approach. The Quantum Insider, which tracks this research closely, reported in late March that the papers collectively represent a step-change in the perceived urgency of Q-Day - the moment a cryptographically relevant quantum computer (CRQC) becomes operational.

There is a second threat that does not require Q-Day to arrive before damage begins. Intelligence agencies and sophisticated nation-state actors are already running what security researchers call "harvest now, decrypt later" operations. They intercept and store encrypted financial communications, intelligence cables, and corporate data today, in bulk, with the intention of decrypting everything retroactively once a CRQC is available. Data with a long secrecy requirement - trade secrets, merger negotiations, long-dated financial instruments, national security communications - is already compromised in transit, sitting in foreign archives waiting for the key that breaks everything.

The Rekeying Problem

9 Years to Replace Every Lock

In August 2024, the National Institute of Standards and Technology published the first three finalised post-quantum cryptographic standards: FIPS 203, FIPS 204, and FIPS 205. These are the replacements. FIPS 203 is based on a mechanism called ML-KEM (Module-Lattice-Based Key-Encapsulation Mechanism), formerly known as CRYSTALS-Kyber. FIPS 204 provides digital signatures via ML-DSA (formerly CRYSTALS-Dilithium). FIPS 205 offers a backup signature scheme, SLH-DSA, based on hash functions rather than lattices - a useful hedge in case the lattice-based approaches reveal unexpected weaknesses.

Lattice-based cryptography works on a fundamentally different hard problem: the shortest vector problem in a high-dimensional lattice. Unlike integer factorisation, no one has yet found a quantum algorithm that efficiently solves it. That "yet" carries weight - but it is the best available option, vetted by NIST over a competitive eight-year evaluation process with global cryptographic input.

The deadlines are firm. RSA, Diffie-Hellman, ECC, and ECDSA will be deprecated by 2030 and disallowed after 2035. The NSA has told vendors of national security systems to begin transitioning immediately, with exclusive use of CNSA 2.0 (the quantum-safe algorithm suite) required by 2030 for most system types. 2026 has been designated the "Year of Quantum Security" by the FBI, NIST, and CISA - a label that signals government seriousness about a deadline that is now inside a single decade.

The practical challenge of migration is not just swapping one algorithm for another. Financial institutions have cryptographic keys embedded in hundreds of systems - HSMs (hardware security modules), payment terminals, certificate chains, API authentication, inter-bank protocols, customer-facing TLS sessions. Most institutions do not have a complete inventory of where their cryptographic assets live. You cannot replace a lock you have not found.

AI on Both Sides of the War

The Tool That Accelerates the Threat - and the Only Tool Big Enough to Solve It

Here is where the story folds back on itself in a way that feels like it was written to unsettle you. AI is not a neutral party in the quantum cryptography race. A Time magazine investigation published in April 2026 documented how AI is actively accelerating the development of quantum computers capable of breaking encryption. Google's and Oratomic's recent results, both with AI-assisted optimisation of quantum circuit design, have significantly shortened the estimated time to a CRQC. The same machine learning techniques that are helping researchers solve the knotty engineering problems of quantum error correction are also compressing the timeline to Q-Day.

But AI is also, simultaneously, the best candidate for the most urgent part of the defense problem - and possibly the only approach that scales.

The first challenge is discovery. CISA published a formal strategy in September 2024 for deploying Automated Cryptography Discovery and Inventory (ACDI) tools across federal civilian infrastructure. The strategy exists because no organisation - not a bank, not a government agency, not a multinational - has a reliable manual inventory of its cryptographic assets. There are too many of them, distributed across too many systems, updated too frequently for human tracking to stay current. AI-powered discovery tools crawl network infrastructure, enumerate every certificate, key, signature algorithm, and encrypted connection, and flag the ones that rely on RSA or ECC. Vendors including Keyfactor, QuSecure, Fortanix, and AppViewX have built these tools specifically for the PQC migration market, which MarketsandMarkets projects will grow from $420 million in 2025 to $2.84 billion by 2030. The money moving into this space reflects genuine urgency from institutions that understand what is at stake.

The second challenge is what the industry calls crypto-agility - the ability to swap cryptographic algorithms without bringing down the systems that depend on them. This concept barely existed as a design principle before the quantum threat made it necessary. Building crypto-agility into financial infrastructure means redesigning systems so that their cryptographic layer is modular: swappable on demand, independently upgradeable, monitored in real time. AI is being deployed to manage the lifecycle of cryptographic assets at scale - shortening certificate lifespans, automating renewals, monitoring for anomalies that suggest a key has been compromised or an algorithm has been deprecated without the system knowing.

The third challenge is real-time threat detection. Research teams are training LSTM (Long Short-Term Memory) neural networks to identify the anomalous patterns that might indicate a quantum-assisted attack in progress - achieving 91 to 99 percent detection accuracy in research settings. The limitation is computational cost: these models are not yet light enough to run in true real-time on production financial infrastructure. That is an active research problem. The detection capability exists in principle; the engineering to deploy it at the scale of a global bank is the next frontier.

The Year of Quantum Security is not a marketing slogan. The FBI, NIST, and CISA attached that designation to 2026 because the migration window is now visibly finite, and because institutions that begin the inventory and planning work now will have a managed transition. Institutions that wait until 2029 will face a chaotic scramble against a hard deadline - and against the rising probability that Q-Day arrives before they are ready.

The April 24 Project Eleven result was 15 bits. Bitcoin uses 256 bits. But in 2019, Google demonstrated quantum supremacy on a task that would have taken a classical supercomputer 10,000 years. In 2024, the qubit threshold to break RSA was 20 million. In early 2026, it is below one million. The curve is not slowing. The padlock on global finance has held for fifty years. The question is no longer whether it can be broken - it is whether we will replace it before someone does.