In 1865 the British Parliament looked at the self-propelled road vehicle and made a decision that would cost the country an industry. It decided the thing was a locomotive. Not a metaphorical locomotive, but a legal one.

The statute was called the Locomotives on Highways Act, and it placed the motorcar in the same legal family as the railway engine, because that was the only shelf in the cupboard that seemed to fit [1]. Everything else followed from the shelf. A railway engine is a heavy, dangerous, crewed machine that runs on a fixed track away from the public. So the law required a crew of three. It capped the speed at four miles an hour in the country and two in town. And it required that a man walk sixty yards ahead of the vehicle carrying a red flag, to warn the world that a locomotive was loose on the Queen’s highway.

Sit with the image, because it is one of the most useful pictures in the history of governance. The machine that would remake the twentieth century, made to crawl behind a strolling pedestrian. A law that looked directly at the future and told it to slow to walking pace.

We enjoy laughing at this. I would like to spoil the joke, because the joke is where the thinking usually stops, and everything interesting lies on the other side of it.

Here is the first discipline of any honest argument. When a story serves your case perfectly, examine it harder than the ones that do not.

The Red Flag Act is the most over-worked anecdote in the whole literature of technology and law, and the version people repeat is not accurate. The flag requirement was watered down in 1878, when the walking distance was cut to twenty yards and the flag itself became a matter of local discretion [2]. The full repeal came in 1896 [3]. And the notion that this single statute is why Britain lost the motorcar to Germany and France does not survive contact with the evidence [4]. Britain’s lag had many parents. Capital that preferred railways and empire to road machinery. An engineering education system that Germany had built better. Road surfaces laid for horses. Fuel distribution that did not exist. Blaming the flag is comfortable because it hands us a villain and a punchline.

So the anecdote fails as economics. It survives, and I think it is close to perfect, as philosophy. Because look again at what actually happened in 1865, before any question of speed or flags arises.

Parliament had to decide what the thing was.

There was no category for it. The vocabulary available in 1865 contained carriages, which were pulled, and locomotives, which were self-propelled. The motorcar was self-propelled. Therefore, it was a locomotive. Therefore, the entire apparatus built for locomotives descended upon it, crew and flag and walking pace, not through malice or even through fear, but through the ordinary force of classification. Once you have decided what a thing is, you have very largely decided how it will be governed. The rest is administration.

That is the argument of this essay, and I will state it plainly now so that everything after it can be tested against it.

The primary act of regulation is not prohibition. It is classification. We do not first decide to restrain a technology and then find words for it. We find words for it, and the restraint follows from the words. And because our words are inherited from the world that is ending, the most consequential decision in the governance of anything new is made at the very moment we understand it least, in a vocabulary that predates it, usually without anyone noticing that a decision has been made at all.

The man with the red flag is not the failure. He is the symptom. The failure happened earlier and more quietly, in the choice of the shelf.

Before going further, I should answer the prior question, because it is less obvious than it sounds and the usual answers are bad.

The Greeks told it as theft. Prometheus takes fire from the gods, gives it to men, and is chained to a rock where an eagle eats his liver each day for eternity [5], [6]. What that myth understands, and what we keep mislaying, is that the gift and the danger are not two things. They are one object. Fire cooks the meal and burns the city with the same indifference. The tool has no view about what it is for.

So, we do not govern technology because it is new. Novelty is not a hazard. We govern it because it is power, and power is indifferent to intention. A hammer does not know whether it is building a house or breaking a skull. A press does not know whether it prints scripture or libel. A distributed ledger does not know whether it is settling an honest trade in Lagos or moving the proceeds of a ransomware attack. The instrument is mute. It waits for a human being.

The purpose of law, then, is narrower and less heroic than its champions claim. It is not to steer the technology. It is to keep the city standing while people learn what they are holding. Law buys time for wisdom, which always arrives late and never arrives on its own.

Which gives us the distinction that governs everything after. The fire is not the arson. A society that cannot tell them apart will spend its history alternating between recklessness and paralysis, and will call both of them prudence.

Law is a technology for encoding yesterday’s consensus. That sounds like a criticism. It is closer to a design specification. Its value lies precisely in its stability, in the fact that a contract signed today will mean the same thing in a court next year, and stability of meaning is purchased with vocabulary that does not move.

Then something arrives that the vocabulary was not built to hold, and the machinery does the only thing it can. It reaches for the nearest available name.

The American test for whether a digital asset is a security still descends from a 1946 case about orange groves in Florida. Investors bought strips of citrus land together with a service contract to farm them, the Supreme Court held that the arrangement amounted to an investment contract, and the reasoning in Howey now determines the legal nature of cryptographic instruments that no member of that court could have conceived [7]. An eighty-year-old judgment about fruit governs mathematics. This is regularly presented as an absurdity. It is not an absurdity. It is the ordinary condition of law, working exactly as designed, reaching for the nearest shelf.

Notice how much is decided by that reach, and how invisibly. Call a token a security and an entire structure descend upon it, built for a world of issuers and shareholders and quarterly disclosure, presupposing that somewhere there is a company with a boardroom and a name on a door. Call it a commodity and a different structure descends, built for grain and copper and oil. Neither vocabulary was written for a thing that has no boardroom, no door and no name, and that may be maintained by six thousand strangers who have never met and are not incorporated anywhere on earth. Yet one of the two must be chosen, because the machinery cannot operate on an unnamed object. And once chosen, the choice hardens into precedent, and the precedent hardens into common sense, and within a decade nobody remembers that a choice was made at all.

Plato worried about this in the Cratylus, asking whether names are natural or merely agreed, and whether a wrong name can mislead us about the thing itself [8]. It is not an idle question for anyone who makes rules. In law, the wrong name does not merely mislead. It legislates.

There is a second difficulty layered on the first, and it is the one most people mean when they say that technology moves faster than law.

David Collingridge stated it in 1980 and nobody has improved on it [9]. When a technology is young, we can still shape it, but we do not yet understand it. By the time we understand it, it has grown so large and so entangled that shaping it becomes surgery on a patient who cannot be anaesthetised. Early on, control without knowledge. Later, knowledge without control. There is no convenient window in which we possess both.

In 1995 nobody could have written sensible law for the internet, because the internet had not decided what it was. By 2015 the platforms had become so woven into commerce, speech and memory that meaningful reform meant unpicking the fabric of daily life. We were never handed the middle moment. We almost never are.

Put the two problems together and you have the real predicament. We must name the thing before we understand it, and the naming is the governing. No legislature invented this problem. It comes with the territory. Anyone who tells you they know precisely how this technology should be regulated is telling you, without meaning to, that they have not understood it.

We flatter ourselves that these are modern difficulties. The oldest surviving code of law contains the answer, and gets closer to it than we usually manage.

Hammurabi ruled Babylon some four thousand years ago. His code has a clause about builders. If a builder builds a house and the house collapses and kills the owner, the builder is put to death [10]. Read it again and notice the silence in it. It says nothing whatsoever about how to build. Not the timber, not the depth of the foundation, not the thickness of the wall. It regulates the outcome and leaves the method entirely alone. Build however you like, invent whatever technique you can imagine, and own every consequence of it.

That is a more sophisticated instinct than most of what we produce today, and it sidesteps the vocabulary problem completely. Hammurabi never had to decide what kind of thing a new building technique was. He only had to decide what a collapse was, and a collapse is a collapse in any century. Regulate the harm and you can govern a technology you do not understand. Regulate the object and you must first name it, and the naming will be wrong.

The Greeks added the second piece. In the Phaedrus, Socrates tells of the god Theuth, who invents writing and presents it to King Thamus as a cure for forgetting [11]. Thamus is unmoved. Writing will weaken memory rather than strengthen it, he says. People will trust marks on a page rather than what they hold in their minds, and will acquire the appearance of wisdom without its substance. Here is the detail that ought to stop us. This is a complaint about a new information technology, made twenty-four centuries ago, and it is very nearly word for word the complaint now made about search engines and artificial intelligence. The anxiety is not new. It recurs with every tool that touches the mind, and it is neither foolish nor decisive. Thamus was partly right. Writing came anyway. The world was not saved and not destroyed. It was changed, which is what technologies do and all they do.

Then the Stoics, who supply the temperament. Heraclitus said you cannot step into the same river twice, because it is not the same river and you are not the same man [12]. Epictetus drew the line that every regulator needs on the wall. Some things are within our power and some are not, and wisdom is knowing which is which [13]. The arrival of a technology is not within our power. Our response to it is. And Solon, asked whether he had given the Athenians the best possible laws, is said to have answered no, he had given them the best laws they would accept [14]. Cynics have been quoting that line for two and a half thousand years. They have it backwards. Solon was describing the beginning of political maturity, not the end of idealism. Law is never the imposition of a perfect design on reality. It is the art of the workable, made by the imperfect, under permanent uncertainty.

Now the objection, and it is a serious one, and any essay that ducks it is advocacy rather than argument.

Sometimes the frightened people are right.

In September 1960 an application landed on the desk of a newly appointed medical officer at the American Food and Drug Administration. The drug was already licensed across Europe. It was marketed as a sedative, widely given to pregnant women for morning sickness, and considered so safe that in parts of Germany it was sold without prescription. The company expected approval as a formality. Frances Oldham Kelsey read the file and found the safety data thin. She asked for more. The company complained to her superiors, called her obstructive and pedantic, and applied every pressure available to it over fourteen months. She would not move [15].

The drug was thalidomide. By the time the pattern of birth defects was understood [16], [17], thousands of children across Europe and beyond had been affected. In the United States, almost none. In 1962 President Kennedy gave Kelsey the highest civilian honour in the federal service, and Congress passed the Kefauver-Harris amendments requiring proof of efficacy and informed consent in trials [18], which is the foundation of modern drug regulation and has saved a number of lives that nobody can count.

Frances Kelsey was the man with the red flag. She walked in front of an innovation, slowed it to a crawl, was told she was standing in the way of progress, and she was right. Any argument about technology and law that cannot accommodate her is not worth reading.

So let us accommodate her properly. What did Kelsey actually do? She did not ban a molecule. She did not declare a class of chemistry illegal or attempt to legislate what a laboratory might synthesise. She stood at one specific place, the point at which the substance would pass from a company into the body of a person who could not assess it, and she demanded evidence before it crossed. She governed the interface, not the invention. And her instrument was reversible. Had the evidence come back clean, the drug would have been approved. She was not making a permanent judgment about a technology. She was holding a door until the facts arrived.

That is the difference between Kelsey and 1865, and it is not a difference of courage or of caution. Both were cautious. One classified the object and governed everything that followed from the name. The other ignored the question of what kind of thing it was and stood at the point of contact with a human being. The second method works on technologies you do not understand. The first cannot.

The record is not all red flags. Four cases deserve attention because they succeeded, and because they succeeded for the same reason, which is not the reason usually given.

The Statute of Anne, 1710. Before it, the Stationers’ Company held a perpetual printing monopoly that functioned as an instrument of censorship. Parliament replaced it with something new. A right belonging to the author, lasting fourteen years and renewable once [19]. The genius of it was the clock. By putting a term on the right, the statute invented the public domain and converted a tool of control into an engine of production. It did not attempt to say what books should be. It regulated a duration.

The Radio Act, 1927. The American airwaves had descended into chaos, with broadcasters transmitting wherever they pleased and drowning one another into uselessness. The response was to license frequencies [20], [21]. This looks superficially like the red flag and is its opposite. The regulators had identified a real structural fact, which is that spectrum is a genuine commons where one person’s use physically destroys another’s. They allocated the collision point and left everything else alone. They did not tell broadcasters what to say or how to build a transmitter.

Asilomar, 1975. When recombinant DNA arrived, the scientists themselves called a voluntary halt, met in California, and agreed a graduated framework of containment tied to assessed risk [22]. As the evidence came in and the early fears proved overstated, the restrictions were relaxed, deliberately and in stages. The moratorium was designed from the outset to be lifted.

The Montreal Protocol, 1987. The most successful environmental treaty ever signed, and it was agreed while the science was still contested [23]. The Antarctic ozone hole had been reported only two years earlier and the causal chain was strongly suspected rather than settled [24]. Rather than wait for certainty, the drafters built the treaty with a ratchet, fixed review points at which the schedule could be tightened as the evidence accumulated. It was tightened, repeatedly, and it worked.

Now look at what these four share, because the common feature is not prudence and it is not foresight.

Not one of them tried to be right.

The Statute of Anne set a term rather than a judgment. The Radio Act governed a collision rather than a medium. Asilomar built a moratorium with a release mechanism. Montreal wrote its own revision into its opening text. The laws that succeeded were not the ones that guessed correctly about the future. They were the ones built to be wrong safely. They had a metabolism. They could take in new evidence and change state without requiring a fresh act of political will each time, which is fortunate, because political will is the scarcest resource in any parliament.

Compare the failures. The Bubble Act of 1720, passed after the South Sea collapse, effectively outlawed the joint-stock company for a century [25], [26]. Parliament saw a genuine abuse and abolished the entire structure that had been abused, taking the instrument of the coming industrial revolution and very nearly throwing it away. It stood, unrevised, until 1825. Compare the licensing of the press, against which Milton wrote the Areopagitica in 1644, arguing that to license thought before it is spoken is to strangle reason in the cradle [27]. Both were permanent answers to provisional questions. Both regulated the fire rather than the arson. Both were, in the language of this essay, category judgments. A decision about what kind of thing this is, hardened into statute, immune to correction.

Now the question directly, and it needs a two-part answer because the single-word versions are both false.

Innovation itself cannot be regulated. Discovery does not ask permission. You cannot pass a law preventing a chemist from noticing a reaction, or a mathematician from seeing a proof, or a developer from realising that a ledger does not require a keeper. The insight occurs inside a human skull, at a location no statute reaches, and it occurs whether or not it is lawful. The history of forbidden ideas is the history of ideas that travelled anyway, usually faster, and usually to somewhere else. You cannot regulate the river at its source. The source is in the mind and the mind is free by construction.

The consequences of innovation can be regulated, and that is not a consolation prize. It is the entire discipline. We cannot stop the discovery. We can determine what it is permitted to do to people who did not consent to it [28], [29]. We cannot legislate the fire. We can certainly legislate the arson.

But I want to sharpen this beyond the usual formula, because “regulate outcomes, not technologies” is a slogan, and slogans do no work.

What is actually governable is the interface. Not the invention, and not even the harm in the abstract, but the specific place at which a technology touches a person who cannot assess it and did not choose it. Kelsey stood at an interface. The Radio Act governed an interface, the point where one signal destroys another. Hammurabi’s builder was answerable at an interface, the moment the roof met the owner. In each case the regulator was able to act without ever answering the question of what kind of thing the technology was, and that is precisely why the rule survived contact with a changing world.

The vocabulary problem dissolves at the interface. A collapse is a collapse. A body is a body. A collision is a collision. These are stable categories in a way that security and commodity and locomotive are not, because they describe events in the world rather than membership of a legal species. And a rule anchored to an event does not need rewriting every time the technology mutates beneath it.

So when I am asked whether innovation can be regulated, I have learned to correct the question gently. Innovation cannot. Consequence must. And the art lies in finding the interface, which is difficult, unglamorous, and almost never what the loudest voices in the room are arguing about.

Sometimes, we do not have to accurately define the thing before we can safely govern it.

In December 2025 the Property (Digital Assets etc) Act came into force and did something English law had not done in centuries [30]. For as long as anyone could remember, personal property came in two kinds. Things in possession, which you can hold. Things in action, which you can only enforce. Every object had to be one or the other. The Act says, in effect, that a thing is not prevented from being property merely because it is neither.

That is worth a second pass, because it is stranger than it looks. Parliament did not define the new category. It did not say what a digital asset is. It said only what it need not be, and left the courts to discover the rest [31]. That is definition by negation, the via negativa of the theologians, arriving at last in property law. And it is the most intellectually honest thing a legislature can do when it knows a thing exists but cannot yet say what it is.

Set that beside the orange grove and you have the two available responses to a new kind of object. One puts it on the nearest shelf and argues about the fit for a decade. The other admits it has no shelf, builds an empty one, and leaves the shape to be found. Neither is obviously right. But only one of them is honest about what it does not know.

Which brings me at last to the American bill, and to the one thing in it that a philosopher should find remarkable. It is not in the headlines and it is not what the lobbying is about.

Buried in the CLARITY Act is a provision permitting a digital asset to change its legal nature over time [32]. A thing may begin life as one kind of legal object, an investment contract with a promoter and a promise, and become another kind, a digital commodity, once the network beneath it has grown sufficiently decentralised. Once no single hand controls it. Once, in a sense, it has grown up and left home.

Consider what that is. It is Heraclitus written into a statute. It is the law conceding, perhaps for the first time in this domain, that things become rather than simply are.

For most of its history law has required objects to hold still long enough to be named. A thing is a security or it is not. A vehicle is a locomotive or it is not. The category is fixed and reality is trimmed to fit inside it. And here, tentatively and with a great deal of drafting awkwardness, is a statute that admits the river flows. That the same token is not the same token twice. That legal identity can mature.

I do not know whether the drafters saw the philosophy in what they were doing, and it does not matter in the least. Whether by insight or by exhaustion, they reached for the one instinct that fits a technology which has not finished becoming. They stopped trying to fix the object and started trying to describe its motion. Set against the whole history sketched in this essay, from the Locomotive Act to the orange grove, that is a real shift in the temperament of law. It is worth more than every jurisdictional boundary in the document, and it will outlast most of them.

Whether it works is a separate matter, and the honest answer is that we do not know. Sufficient decentralisation is a threshold, thresholds invite gaming, and a great deal will depend on who measures and how often [33]. But the instinct is sound, and the instinct is the part that transfers to the next technology and the one after that.

And then there is the name. They called it the Clarity Act. There is a confession in that. You do not name a law after the virtue you already possess. You name it after the one you lack. To call a statute Clarity is to admit in the title that everything preceding it was fog, which is a more Socratic act than anything in the surrounding debate. The beginning of wisdom is the acknowledgement of what one does not know, and it is not often that a legislature manages it on the cover page.

As I write this essay in July 2026, the thing is still not law. The Senate committee moved it in May. It may pass soon or it may slip past the midterms into another Congress and another vocabulary. There is a lesson in that grinding slowness as well. Law moves at walking pace while the technology moves at the speed of light, and the gap is not the fault of any particular senator. It is the permanent condition. It is the man with the red flag, walking his sixty yards ahead, while behind him the engine strains against four miles an hour.

I am a physician as well as a blockchain professional, and I want to close on the one professional habit that I believe transfers directly. Medicine has been governing dangerous power under radical uncertainty for a very long time and has made every mistake in this essay at least twice, including the naming mistake.

For the better part of a century, neurology and psychiatry carried a diagnosis called hysteria. A category inherited from antiquity, built on a wrong idea about the female body, applied to patients with real and disabling neurological symptoms that could not be traced to a lesion. The observations were not wrong. Those patients were genuinely unable to walk, or to see, or to speak. What was wrong was the shelf they were placed on, and the shelf determined everything that followed. It determined who treated them, whether they were believed, whether they were suspected of malingering, and whether anyone bothered to investigate the mechanism. The category did more damage than the disorder. It took until well into this century for the field to reclassify the condition as a functional neurological disorder, a problem of nervous system functioning rather than of character [34], [35], and the reclassification changed the treatment, the research funding and the dignity of the people who had it. Nothing about the patients changed. The name changed, and with it, everything.

That is the ontology problem in a clinic rather than a parliament, and it is why I take it seriously in both. Get the category wrong and you can be humane, diligent, well intentioned and busily gathering evidence, and still do harm for a hundred years.

The rest of the clinical posture follows. A doctor may never have complete information. The imaging is ambiguous, the history is partial, the literature is unsettled, and the patient is in front of you this morning rather than in some future decade when the evidence has matured. You do not get to wait for certainty. You never had that option. So, you learn a stance that is neither the coward’s paralysis nor the tyrant’s confidence. You reason from the best evidence available. You intervene in proportion to the risk you can actually demonstrate, not the risk you can imagine. You take the smallest effective step. You write down what you expect to happen so that you will know if you were wrong. And you hold to the oldest instruction in the trade, primum non nocere, first do no harm, which does not mean do nothing. Those are entirely different instructions and confusing them has killed people in both directions.

Read the four successes again with that in mind and you will find the same posture. Montreal is a treatment plan with review dates. Asilomar is a precaution written with the intention of being withdrawn. Neither pretended to certainty. Both were built to learn.

Regulation done properly is not a monument you erect once and photograph. It is a treatment you adjust as the patient responds. The men of 1865 failed this in every respect. They classified on the basis of the nearest available word, they intervened everywhere rather than at the point of contact, they took the largest step available rather than the smallest that worked, and they did not revise for thirty-one years while the future walked past them at four miles an hour.

The Act was repealed in 1896, and they celebrated with a drive from London to Brighton to mark the raising of the limit and the removal of the walking man. They called it the Emancipation Run. It is still held every November, a procession of the oldest cars in Britain moving down through Surrey and Sussex, commemorating the day the law finally stopped walking in front of the future.

I think about that procession whenever I read the debates around the CLARITY Act and every act that will follow it. Somewhere ahead of each new technology there is still a man with a red flag, walking his careful sixty yards, sincerely frightened and entirely reasonable and, quite often, wrong in a way that will only be obvious to people not yet born. The task is not to abolish him. Sometimes he is Frances Kelsey, and a society with no red flag at all is a society that finds out about thalidomide the hard way.

The task is to teach the red flag man where to stand. Not in front of the fire but beside the arson. Not at the workshop door but at the interface, the place where the thing touches somebody who did not choose it and cannot assess it. And to teach him to write his rules the way a clinician writes a prescription, with a dose, a review date and an honest expectation of being wrong.

We have had the material for a long time. Hammurabi knew to regulate the collapse rather than the blueprint. Thamus knew that the fear of a new tool is older than any particular tool. Montreal knew that a law which cannot learn is a law that will need to be broken. The CLARITY Act, in its better moments, knows that a thing can change its legal nature as it grows, and admits in its own title that what came before it was fog.

The one lesson we have never managed to hold is that all of this must be remembered at the moment it is hardest to remember, which is always now, and always in the dark, with the engine straining behind us and the flag still in our hand. The wise society is the one that finally works out where to set it down.

_______________________________________________________________