Digital Assets and Tokenisation: Categories, Custody and Control

A digital asset here is a claim or an item of value recorded as a token on a shared record, where the record is the asset rather than a description of one. Four categories are what a regulated Indian lender actually meets. Control of the asset is control of a private key, and somebody holds that key: either the customer or an institution acting for them.

Every question ever handed to an analyst about one of these collapses into two ordinary questions, and both are far older than anything digital. The first is the claim: what is it, and on whom? A token is worth exactly what its issuer owes and not a paisa more. The second is the key: who holds it? The ability to instruct is the whole of control, and it sits with exactly one party at any given moment, whatever the arrangement is called in the brochure. Answer both, and every arrangement on offer resolves into a claim on a named debtor and a key in a named pair of hands.

What is a digital asset in the sense a lender actually meets?

The cloakroom of a wedding hall is where the whole subject is already standing, and nobody in the queue calls it anything. A guest hands over a shawl and is handed a small numbered ticket. Nobody believes the ticket is the shawl. And yet everybody in that queue understands three things about the ticket without ever being told them. The ticket is a claim on the cloakroom and on nobody else in the building. Presenting the ticket is how the shawl moves. Whoever holds the number walks away with what the number points at. And if the ticket is dropped in the car park on the way out, the guest has a problem that no amount of explaining who she is will fix quickly.

The claim, the transfer and the loss are the entire subject. The only thing that changes when the setting becomes a bank is that the ticket stops being cardboard. A digital assetA claim or item of value recorded as a token on a shared record, where the record is the asset itself., in the sense a regulated lender actually meets one, is a claim or an item of value recorded as a token on a shared record, where the record is the asset rather than a description of one.

The last clause is worth slowing down for. Most of the confusion about digital assets lives there. A passbook is a record about a deposit. The deposit itself sits in the bank's own ledger. The passbook was never the thing, so losing it in a house move leaves the deposit entirely untouched. A token on a shared record is the other arrangement altogether. There is no separate original sitting elsewhere that the token is copying and reporting on. The entry is the claim.

A digital asset is not a new kind of value. A digital asset is an ordinary claim on an ordinary debtor, kept in a place where the record itself is what moves. Every property that people find strange about the subject, including the ones that make lawyers unhappy, follows from that single relocation and from nothing else.

What does tokenisation actually change?

Tokenisation: the record becomes the thing that moves

A kirana shop keeps a notebook for credit. A customer's name sits in it with a balance against it. Suppose she wants her neighbour to have the credit she is owed for the crate of bottles she returned. She can tell him about it. She can write him a signed note about it. She can photograph the entry and send it to him. None of that moves anything at all. The credit moves at exactly one moment: when the shopkeeper strikes her name out of the notebook and writes his. The notebook is not a description of who is owed what. The notebook is who is owed what.

TokenisationRecording a claim as a token so that moving the token is the transfer rather than a message about one. takes that observation seriously and builds on it. A claim is recorded as a token on a shared record, and moving the token is the transfer, rather than a message announcing a transfer that somebody else will complete somewhere else later.

Do that, and a journey with five moving parts becomes a journey with three. On the older path there is an instruction, then a message carrying it, then a settlement step where the two sides actually change places, then a confirmation coming back, then a reconciliation to check the confirmation was true. On a shared record the holder instructs using their own key, the record accepts the entry, and the claim is with the recipient. Nothing is left to confirm, so no confirmation follows. There is one record and nothing to reconcile it against, so no reconciliation follows either.

Tokenisation is not making a record digital. Tokenisation makes the record the place where the transfer happens, and the instruction and the settlement stop being two separate events with a working day sitting between them.

How that shared record is built, what kinds of network it can run on and what an automatically executing agreement sitting on one does are set out under blockchain and distributed ledger. The subject here begins one step later, at what is being recorded and who can move it.

Is tokenisation just digitisation with better software?

Tokenisation vs Digitisation, and why the difference is not one of degree

The two words name changes to different parts of the arrangement, and the difference deserves a test rather than another definition.

DigitisationMaking an existing record readable by a machine, leaving the transfer exactly where it already was. is making an existing record readable by a machine. A bank that scans forty years of paper certificates, indexes them and puts them behind a search box has done something genuinely useful. Staff stop walking to a strongroom. Nothing is lost to damp or to a misfiled folder. A request that used to take a day takes a minute. And the transfer of a certificate afterwards happens exactly the way it happened before: an instruction goes somewhere, a person or a system somewhere else amends a register, a confirmation comes back, and somebody reconciles the two.

The test is one question and it is not a question about technology: after the change, is moving the record itself the transfer, or does the record still describe a transfer that happens somewhere else? Digitisation answers that question no. Tokenisation answers it yes.

There is nothing in between. Calling tokenisation an advanced form of digitisation is what causes trouble later on. Digitisation and tokenisation are not two points on one scale of modernity. The two change different things. One changes how a record is read, and the other changes where a transfer happens. A bank will very often do both, in that order, on the same asset, and finishing the first does not put it part of the way through the second any more than tarring a road is part of the way to building a railway.

Which categories does a regulated Indian lender actually meet?

Four, and it is worth being blunt about why a numbered list of four beats a general definition here. A general definition invites sorting by how these things are built, and that is the least useful axis available. Sorting them by who owes the holder something puts the risk in plain sight in the first second, before anybody has said a word about the record.

The four below are the ones the bank in this case examined, and every figure attached to them is that bank's own. The numbering makes it possible to point at one of the four without ambiguity.

The four are not four technologies. The four are four different debtors, and every other property worth caring about follows from that one column.

What is each category a claim on, and who owes it?

Category 1, a tokenised depositA claim on a bank recorded as a token on a shared record., is the one the bank in this case actually used. A deposit at a bank was always a claim on that bank, and it was always worth exactly what that bank could pay. Recording it as a token changes how it travels between holders and changes nothing whatever about who has to pay it back on the day it is asked for.

Category 2, a tokenised securityA token that records ownership of a security., is a token that records ownership of a security. The obliged party is the issuer of that security, and everything that security always carried, good and bad, it still carries. Category 3, a central bank digital currencyA digital representation of value issued by the central bank., is a digital representation of value issued by the central bank, so the obliged party is the central bank itself.

Category 4 is the one that catches people, and it is worth an extra minute. A token that records ownership of a physical or contractual asset, a warehouse receipt over sacks sitting in a shed, or an invoice a buyer has not yet paid, brings in a debtor the first three do not have: somebody has to actually deliver a physical thing or pay a real bill. The record can be immaculate while the shed is empty. A shared record cannot see inside the shed and never claimed it could, so every check that used to be done on the shed still has to be done on the shed.

The token changes how a claim moves and changes nothing at all about who owes it. The first question to ask of any of the four is therefore the oldest question in credit, not anything about the record.

What sits outside these four categories, and who states the Indian position?

A whole area sits next door to these four categories, and the line between them is worth drawing plainly.

Publicly issued digital tokens traded on open markets are a separate subject, and any Indian position on them is stated by the Reserve Bank of India and the Securities and Exchange Board of India. Both bodies publish at rbi.org.in and sebi.gov.in respectively.

A summary of a position goes out of date the moment the position moves, and anybody who learned the summary carries a wrong thing around for years without knowing it. The address of the body that states the position does not go out of date. Naming the source is the answer to that question, not a way of avoiding it.

What is a private key, and why is it not a password?

Private Key: the whole of the ability to instruct

A password-shaped idea is what almost everybody brings to a private key, and the idea is wrong in a way that matters more than anything else in the subject.

A password is a secret sent to somebody. There is an institution at the other end holding something it can check that secret against, and precisely because that institution holds it, that institution can change it. The forgotten password link exists because there is an issuer sitting behind the whole arrangement. Nothing is really being proved to the door. The holder is asking an organisation that already knows him to open it, and if he forgets the words, the organisation has other ways to satisfy itself and will simply issue new ones.

A private keyThe value whose possession is the whole of the ability to instruct, and which cannot be reset or reissued by anybody. works the other way round. A private key is a value that is never sent to anybody at all. The key produces an instruction that could only have been produced by whoever holds it. There is nobody at the other end playing the checking role, and so nobody at the other end keeps a copy to compare against. There is no issuer of a private key, and so there is no issuer to ask.

A password is a request to an institution to act on the holder's behalf. A private key is the act itself, and that is precisely why nobody can hand out a new one.

One sentence carries the rest of the subject, and it is worth reading twice. Whoever holds the private key controls the asset. Not is entitled to it. Not can prove they are the holder of record. Controls it: can move it, right now, without asking anybody's permission and without anybody being in a position to refuse.

A signature made with such a key raises two further questions, what it proves about who signed and what it takes for that proof to still stand years later, and both are set out under digital signatures. The key on its own is only what it gives whoever is holding it.

What does a digital wallet actually hold?

Digital Wallet: what is actually inside it

The name is the problem, and it has probably done more damage to public understanding of this subject than any other single word. A wallet in a pocket holds money. So the phrase invites a reasonable assumption that something of value has been put inside something on a handset.

A digital walletSoftware that holds keys, and never the asset itself. is software that holds keys. A wallet never holds the asset. The claim sits on the shared record. The record is somewhere else entirely and is not inside anybody's phone.

The household version runs the same way. There is a keyring in a kitchen drawer and one of the keys on it opens a bank locker. The jewellery is not in the drawer. If the keyring is lost on a bus the jewellery has not moved a millimetre; what has been lost is the ability to reach it. A bank stands behind the locker and can be asked, with enough patience and paperwork, to open it another way. In the key case there is nobody to ask, unless somebody else is already holding the key, and who that somebody is turns out to be the whole custody question.

A wallet holds keys and never the asset, so the question where is my digital asset has no answer that points at a device. One useful consequence follows immediately. How many wallets a key sits in is not the interesting question. How many places that key value exists is the interesting question. A value can be copied onto paper, held in two programs at once, or written into a note on a laptop by somebody being helpful. Every one of those places is a place where control now sits.

Who controls the asset, then?

Whoever holds the private key. The answer stops there and does not come with a second clause.

CustodyThe arrangement that answers who holds the private key, which is the same question as who can act. answers who is holding it. Two arrangements are on offer, numbered so that either can be named without ambiguity. Arrangement 1, the customer holds the private key. Arrangement 2, an institution holds it for the customer. Two is the entire menu, and the reason there are two rather than a dial of many is worth stating carefully.

A spare key to a front door goes to a neighbour before a family travels. While she has it, she can open that door, and there is no arrangement of words that makes this less true. The neighbour's ability to open the door is not a statement about entitlement or trust or paperwork, but a statement about who can physically open the door this evening, and the whole of this subject rests on that one statement.

There is no middle. At any moment exactly one party can produce a valid instruction, and that party controls the asset whatever the arrangement is called on the statement. Arrangements described as shared or joint or split do exist, and the way to read any of them is always the same: which party can produce a valid instruction right now without the other party taking part. Whichever party that is, control actually sits there, however the paperwork describes it.

What does each arrangement trade away?

Two conditions test them, and the tidy, uncomfortable finding is that each arrangement passes one condition and fails the other. Neither is a strong version and a weak version of the same thing.

Condition A is that the customer loses the key. Under arrangement 1 that is the end of it. There was never an issuer holding a copy, so the ability to instruct is gone permanently, and no institution, court or engineer can restore it. Under arrangement 2 the same event is a support case: an identity check, a form, a wait, and access comes back. For an ordinary retail customer, most of whom have lost a password at some point in the last year and thought nothing of it, that is a real and serious improvement rather than a cosmetic one.

Condition B is that the institution cannot act. Under arrangement 1, nothing happens to the customer at all; they were never depending on it. Under arrangement 2 the customer cannot act. People usually think of only the first part of that dependence, so it is worth breaking the dependence into three. The institution must continue to exist. The institution must remain able to act, and remaining able is a different thing from existing. And it must choose to act, a different thing again.

Neither arrangement removes the failure. Each one chooses which failure it will have, and the only honest way to write the choice down is to name both halves of it. Easier to explain is not the same as better. A support case is far easier to explain to a customer standing at a counter than a permanent loss is, and that fact says something about which conversation a firm would rather have, not about which arrangement is stronger in general.

What did one bank choose for its 1,200 pilot holders?

Sumeru Bank Limited ran a pilot on category 1, a tokenised deposit, recorded on a shared record it ran together with four other institutions, five in all. Over three months the five put 1,200 instructions across it, and settlement between them fell from 1 working day to under a minute.

Three readings sit on top of those counts, and all three are arithmetic rather than measurement. First, 1,200 instructions over three months averages 400 a month. Second, 1,200 instructions against 1,200 holders averages one instruction per holder. Nobody knows the distribution behind either average. Third, this bank's own working day is 7 hours, being 420 minutes, and a fall from 1 working day to under a minute removes more than 99.7 per cent of the elapsed time, a factor of more than 420.

One coincidence is worth killing on sight. The intake chain's own shadow run also lasted three months at about 400 files a month. The two are different pilots on different things, and the matching monthly rate is arithmetic rather than a relationship. Nobody should ever put the two counts in one sentence.

On the custody question the bank chose arrangement 2 for all 1,200 pilot holders, so an institution held every key. Inside the bank the choice was written down as removing a risk, and that wording is where the trouble starts.

Why these two structures are categorical rather than continuous

The two structures do not admit of degrees, and that is a fact about the subject rather than a simplification of it.

An asset is one of four categories, and a key is held by the customer or by an institution. A control that slid from arrangement 1 towards arrangement 2 would paint a spectrum where no spectrum exists, and suggest a setting somewhere in the middle at which control is partly here and partly there.

A dial between the two arrangements would offer a setting that custody has no room for. At any moment exactly one party can produce a valid instruction, and there is no fractional version of that. The comparison is drawn instead, with the trade written on both sides. The pilot's counts are one bank's own figures, and they do not move.