No midnight

I remember Y2K

At Coopers & Lybrand our engagement letters were very standard. Then, some years before the millennium, we were told to add a disclaimer about Y2K, the millennium bug. It said, in effect, that we could not be held responsible if we missed a Y2K issue: that was management’s responsibility.

At first there was no extra fee income. Then a select band of IT risk specialists started to take on Y2K work. The engagements never seemed very interesting to me, although in retrospect they were really important. As 1999 approached, there was clearly a growing feeding frenzy.

From 31 December 1999 into 1 January 2000 I was glued to the news, listening for Y2K stories. No aircraft fell from the sky. Washing machines kept washing, pay reached bank accounts and mortgage payments went out as normal. A total damp squib. A few anecdotes circulated later, but the general view was that there had been a panic about nothing. Or perhaps the panic was avoided by the diligent work of IT departments and consultants around the world.

Quantum computing now tugs at my mind in the same way. People tell me that they keep hearing the word, without anyone explaining why it matters. It has at least as much potential for damage as Y2K, perhaps more, because banking as we know it depends on the encryption at risk. And it is harder, for one simple reason: there is no midnight.

Quantum in plain words

An ordinary computer works in bits, each a 0 or a 1. A quantum computer works in qubits, which follow the rules of quantum physics. For most tasks that gives no advantage. For a few particular kinds of problem, a large enough quantum computer would be vastly faster than any ordinary machine.

One of those problems sits under much of today’s security. The encryption that protects online banking, card payments, secure email, digital signatures and software updates relies on certain mathematical problems taking an impossibly long time to solve. The best known are called RSA and elliptic curve cryptography. In 1994 a mathematician, Peter Shor, showed that a large quantum computer could solve those problems quickly. That would undermine the methods used both to set up secure connections and to verify digital signatures. Confidential information could be exposed, and signatures forged.

No such machine exists yet. Today’s quantum computers are too small and make too many errors. Security people call the machine that would break today’s encryption a cryptographically relevant quantum computer.

Two points of balance. Not all encryption is equally exposed: the kind used to scramble stored data and the fingerprints used to check files are far less affected, and longer keys are generally enough. And there is a fix. Post-quantum cryptography means new methods that quantum computers are not known to break. The US standards body, NIST, published the first three standards in August 2024. The problem is not inventing the fix. It is replacing the old methods wherever they are used in your tech estate.

No midnight

Y2K had a precise date. Everyone knew when the risk would materialise, and on the night everyone could see whether the work had held. Quantum has neither.

Nobody knows when a cryptographically relevant quantum computer will exist, and estimates of the machine needed keep falling. In 2025 a Google researcher put the machine needed to break the standard RSA key at under a million qubits. In late October a French start-up, Alice & Bob, estimated that its own design could do it with about 22,600 qubits, in about three days. That is a company’s projection for hardware it has not yet built, in a paper not yet peer reviewed. The estimate is not a prediction that such a machine will exist. But it illustrates how rapidly the engineering assumptions are changing. The first organisation to get there may not announce it.

The harm also does not wait for the day. An attacker can copy encrypted data now and keep it until it can be read. Security people call this “harvest now, decrypt later”. So the question is not only when the machine arrives. It is how long your data has to stay secret. Customer records, health data, merger papers and long-dated contracts may still matter in 2035. If they are copied today, some of the damage may already be done, and nobody will know until much later.

There will be no night to stay up for, and no morning to check.

The damp squib problem

Y2K looked like a panic about nothing. Much of that was because the work had been done. But once the deadline passed without catastrophe, it became easy to question whether all that expenditure had been necessary. That is the trap. When preparation succeeds, it leaves no evidence that it was needed.

Quantum migration will be worse in this respect. It will take years, cost real money and compete for budget with projects whose benefits are visible. If it works, nothing will happen. There is no deadline to concentrate minds, and no night when the board can see the result. Every year it will be easy to argue that the machine is still some way off and the money is better spent elsewhere.

So this is a question of risk appetite, not simply technology. The board has to decide how much exposure it will accept, for how long, and what would make it act sooner. It has to keep funding work the success of which will look like nothing, and accept that the people who did it well may never get the credit.

One other memory from the 1990s is worth keeping. Our disclaimer moved responsibility for Y2K firmly onto management. Expect the modern version: suppliers’ contracts that leave quantum safety with you.

What a board can do now

The lasting legacy of Y2K was not the night. It was that firms finally learned what systems they had. Quantum needs the same discipline, applied to encryption.

The FCA has now said so too. On 4 November it told financial firms to act now: to understand the cryptography they use, set risk-based migration objectives and carry their plans through. “Senior managers should ensure that their firms allocate enough time, investment and resources to manage PQC migration safely and meet the relevant deadlines.” It added that failed change projects accounted for around a fifth of all incidents reported to it in 2025. The migration is itself a risk to manage.

1. Decide what must stay secret longest. Which data would still do harm if read in ten or fifteen years? That sets how urgent the work is.
2. Ask for an inventory. Where does the business rely on public-key encryption, what does it protect, and who controls it? Much of it sits inside suppliers’ products and services.
3. Ask suppliers for their plans. When will their products use post-quantum methods, and what do the contracts say if they don’t?
4. Set the plan against the timetable. The UK’s NCSC asks organisations to complete discovery and an initial plan by 2028, their highest-priority migration by 2031 and full migration by 2035. For financial services, the G7 Cyber Expert Group’s voluntary roadmap points to the most important systems around 2030 to 2032.

None of this needs the board to understand qubits. It needs the board to keep asking, year after year, while nothing appears to happen.