Author: Dimitar Kolev*

A transfer leaves a public blockchain at 10:14. The custodian’s internal ledger still shows the sender’s balance at 10:16. A bridge records the transfer as initiated but not yet completed. The receiving platform credits the destination account at 10:21, while its settlement report treats the position as final only after an additional confirmation cycle. None of these records is necessarily wrong. They may be recording different stages, functions or legal relationships. Yet in arbitration the temptation is often to ask which record is the “real” one.

That question is usually too crude. A public blockchain may be the best evidence that a transaction was included in a block. A custodian’s ledger may be the better evidence of what it owed to a client at a particular time. A bridge record may show whether a cross-chain process reached the stage required by the protocol. A corporate or token register may speak to an entitlement that the blockchain itself does not determine. Conflicting records therefore do not always present a contest between truth and error. They may present a problem of function, timing, legal relationship and the proposition actually in issue.

This post proposes a proposition-specific approach. Before deciding which record deserves greater weight, a tribunal should ask what fact each record is offered to prove, what system generated it, what that system was designed to record, at what reference time the record is reliable, whether the record is complete and final, and how it reconciles with other technical and legal materials. The task is not to identify a universally superior system. It is to determine which record answers which question.

Different Systems Can Be Correct at the Same Time

Digital-asset transactions often pass through several technical and legal layers. The public chain may record movement between addresses. A custodian may maintain an omnibus wallet while allocating client positions only on an internal ledger. An exchange may treat a trade as executed before withdrawal is available. A bridge may record lock-up on one network before minting or release occurs on another. A tokenised asset may have an on-chain representation while legal title or entitlement depends on an off-chain register, issuer record or contractual condition.

The apparent contradiction may therefore result from the systems answering different questions. One record may concern technical execution, another account allocation, another settlement, and another legal entitlement. A discrepancy can also be temporal. Blockchain finality, platform accounting, bridge completion and contractual cut-off times may not coincide. The same event can be “complete” for one system and still provisional for another.

The point is visible in litigation over failed crypto platforms. In In re Celsius Network LLC, 647 B.R. 631 (Bankr. S.D.N.Y. 2023), the court treated ownership of assets in Earn Accounts as a contract-law question and held, on the unambiguous Terms of Use, that the deposited assets became property of Celsius and then of the bankruptcy estates — while expressly reserving individual account holders’ defences to contract formation for a later stage. In Ruscoe v. Cryptopia Ltd. (in liq.), [2020] NZHC 728 (N.Z.), the New Zealand High Court reached a materially different proprietary result: cryptocurrencies were property and were held on trust for the relevant account holders. Crucially for evidentiary purposes, the court relied on Cryptopia’s internal SQL database to establish certainty of subject matter and of objects, identifying both the currencies and the account holders’ recorded balances. The platform held every private key; the public chain and that custody did not by themselves answer the proprietary question. The internal platform record supplied part of the legally decisive identification.

Start With the Proposition, Not the Technology

The first procedural question should be: what proposition is this record offered to prove? A transaction hash may prove that a transaction was recorded. It may not prove who had legal authority to cause it. An internal ledger may prove how a platform allocated a balance to a client account. It may not prove where the underlying assets were held. A registry entry may prove a legally recognised status under the applicable law while saying little about who exercised technical control over a wallet.

This matters because parties often use broad words such as “ownership”, “control”, “settlement” or “transfer” to describe different things. If the tribunal does not define the proposition precisely, it can end up comparing records that are not actually inconsistent. A blockchain record and a custody ledger may appear to disagree only because one records an asset movement and the other records a contractual account position.

A useful case-management step is therefore to require each party relying on a system-generated record to identify the proposition for which it is relied upon. That simple discipline narrows the dispute. It also exposes situations in which the evidentiary disagreement is not about authenticity at all, but about the legal meaning attached to a technically accurate event.

The same discipline appears in tracing cases. In the ex parte proprietary-injunction application AA v. Persons Unknown, [2019] EWHC 3556 (Comm) (Eng.), the claimant’s investigator traced 96 of the 109.25 Bitcoin paid as ransom to an address associated with the Bitfinex exchange; the balance had been converted into fiat and dissipated. The chain analysis could establish movement to the address. It could not identify the person behind it. Bryan J proceeded on the basis that the identifying information was held, or likely held, by the exchange under its know-your-customer obligations. The point is evidentiary rather than a final merits finding: the public-chain analysis and the platform records were capable of proving different propositions, and the latter were needed to move from an address to an identifiable account holder.

Where the Reconciliation Logic Comes From

The framework is not derived from a hierarchy of digital-asset records. It begins with ordinary arbitral evidence logic. Article 9.1 of the 2020 IBA Rules on the Taking of Evidence places admissibility, relevance, materiality and weight with the tribunal. Those concepts are proposition-dependent: a record cannot be weighed sensibly until the tribunal knows what it is supposed to prove. The IBA framework prescribes no digital-asset test, but it supplies the starting discipline — define the evidentiary question before assessing the source.

A second line comes from private-law treatment of digital assets. In the 2023 UNIDROIT Principles on Digital Assets and Private Law, Principle 6 defines control as a factual concept. The commentary to the transfer principle separately makes clear that a change of control is not the same thing as a transfer of proprietary rights, while custody is treated separately in Chapter 4. Celsius and Cryptopia show why that separation matters in practice: technical custody and account records operate within legal relationships that may produce very different proprietary consequences.

A third line concerns time and finality. Principle 8 of the CPMI-IOSCO Principles for Financial Market Infrastructures requires an FMI’s rules and procedures to define clearly the point at which settlement is final, final settlement being the irrevocable and unconditional transfer of an asset or discharge of an obligation. The CPMI’s analytical framework for distributed ledger technology in payment, clearing and settlement likewise treats DLT as one part of a wider settlement arrangement rather than assuming that ledger recording alone answers every settlement question. Taken together, these sources point toward function, provenance, reference time, completeness, finality and legal effect as overlapping questions for reconciliation.

A Six-Factor Reconciliation Framework

Those overlapping questions can be organised into six practical factors. They are not a checklist in which each factor has a separate source or fixed weight; they are a synthesis for deciding why two records appear to conflict and what each can actually establish.

First, system function. What does the system actually record? A public chain may record transactions and state changes; an exchange ledger may record client entitlements; a bridge may record protocol stages; a corporate register may record a legally recognised interest. Reliability cannot be assessed independently of function.

Second, provenance. How was the record generated, extracted and preserved? Native system data, independently reproducible on-chain data and platform exports may carry different evidentiary features. Screenshots can orient the tribunal but may omit surrounding context, metadata or later corrections.

Third, reference time. Which moment matters? A blockchain timestamp, platform server time, contractual cut-off, settlement time and finality point may differ. A party can appear to have had a balance at one reference time and not another without either record being false.

Fourth, completeness. Is the record the complete source or only one layer of a process? A bridge event on the source chain may be incomplete evidence of destination-chain completion. An exchange trade confirmation may not establish that assets became withdrawable or finally settled.

Fifth, finality and reversibility. Was the event provisional, reversible or subject to challenge, reconciliation or administrative correction? Finality is not simply the number of confirmations displayed by a system. In settlement infrastructure it is a defined point at which transfer or discharge becomes irrevocable and unconditional under the relevant rules and legal framework. A tribunal should therefore identify whose concept of finality matters to the claim before it.

Sixth, legal effect. What does the contract and applicable law make of the technical event? A record can establish that something happened without establishing that the event discharged an obligation, transferred title, satisfied a condition or bound the party said to be responsible.

Three Recurring Conflict Patterns

Consider first the difference between a public-chain balance and a custodial account balance. A custodian may hold assets in an omnibus address while maintaining client positions internally. The public chain may accurately show the custodian’s aggregate holdings, while the internal ledger accurately shows the client’s contractual position. Neither record should automatically displace the other because they answer different questions.

Second, consider a cross-chain bridge. A source-chain transaction may be irreversible while the destination-side process has not completed. If the dispute concerns delivery on the destination network by a contractual deadline, the source-chain record may be important but not conclusive. Protocol documentation may establish the mechanics of the bridge, but the legal consequence still depends on the contract, the relevant reference time and the applicable law.

Third, automated execution itself may require legal reconciliation. In Quoine Pte. Ltd. v. B2C2 Ltd., [2020] SGCA(I) 2 (Sing.), algorithmic trades were automatically executed and settled on a cryptocurrency exchange at roughly 250 times the prevailing market rate after an operational failure; the platform reversed them the following day. The court still had to analyse the contractual consequences of the automated trades and the doctrine of unilateral mistake, and held the reversal to be a breach of contract. The system’s execution record established what the platform did. It did not, by itself, answer whether the resulting contracts were binding or reversible.

The Tribunal Should Reconcile, Not Rank

The authorities above point away from a general hierarchy of system-generated evidence. Public blockchain data may be highly reproducible and resistant to alteration, but that does not make it legally complete. Internal records may be editable, but editability does not make them unreliable where the system’s function, audit trail, controls and surrounding evidence support them. Weight follows the proposition and context, not the label attached to the technology.

Where authenticity, completeness or reconciliation is disputed, targeted contextual production may be more useful than a broad demand for entire databases or technical environments. The tribunal may ask for surrounding entries, audit logs, extraction methodology, relevant system rules or an expert comparison limited to the disputed point. The objective should be to make the records testable without turning a focused evidentiary issue into an uncontrolled technical inquiry. That approach is consistent with the risk-based logic of the ICCA-NYC Bar-CPR Protocol on Cybersecurity in International Arbitration, which favours proportionate measures over a single technical solution.

Experts Can Explain the Systems; They Should Not Decide the Legal Conflict

Expert evidence may be necessary where different systems use different data structures, timestamps, settlement mechanisms or reconciliation processes. An expert can explain what each record means technically, whether the extraction is reproducible, what assumptions are required and whether two apparent discrepancies are in fact compatible.

But the mandate should stop short of asking the expert which party “owned” the asset, whether contractual performance occurred or which record has legal priority. Those questions depend on the arbitration agreement, the contract, applicable law and the tribunal’s assessment of the evidence as a whole. Technical reconciliation should support adjudication, not replace it.

The Limits of the Approach

Proposition-specific reconciliation is not costless. It front-loads work into the early procedural phase, at a point when the parties’ cases may not yet be fully formed, and it asks tribunals to make characterisation decisions before the evidentiary record is complete. It can also be used tactically: a party that frames its propositions expansively can justify wide requests for surrounding entries, audit logs and system rules under the banner of contextual testing, converting a narrow dispute into a broad technical inquiry — precisely the outcome the approach is meant to avoid. The discipline therefore depends on tribunals policing proportionality as firmly as they police relevance. Nor does the framework resolve genuine conflicts of law: where two systems are each accurate within their function but the applicable law attaches decisive weight to one, reconciliation clarifies the question without answering it.

One Procedural Illustration

The TAC Digital Assets & Blockchain Disputes Supplementary Rules provide one procedural illustration of this approach where the parties have expressly selected them. Article 6 directs attention to the integrity, source, extraction method, completeness, provenance, limitations, reproducibility and testability of system-generated records, while allowing surrounding or contextual records where a disputed record cannot be understood in isolation. The Rules also separate technical records and control from legal ownership, authority and entitlement. They do not create a hierarchy in which blockchain evidence automatically prevails over platform or other records.

Conclusion

In digital-asset arbitration, the wrong question is often “Which system is right?” Different systems may be right about different things. A blockchain may accurately record execution, a custodian may accurately record an account position, a bridge may accurately record an intermediate state, and a registry may accurately record a legal entitlement. The conflict appears only when those records are asked to prove more than their function supports.

The better approach is proposition-specific reconciliation. Define the fact to be proved; identify the function and provenance of each system; fix the relevant reference time; test completeness and finality; reconcile competing records; and only then determine legal effect. The tribunal’s task is not to choose a universally superior record. It is to explain why a particular record proves a particular proposition in the legal dispute before it.

*Dimitar Kolev writes on international arbitration, digital assets and dispute resolution, and is affiliated with the Ticino Arbitration Centre (TAC)