Should a Tribunal Ever Order Production of a Private Key?
November 16, 2026
An arbitral tribunal receives a request to produce the private key said to control a wallet. The requesting party maintains that the key is the best evidence of control. The respondent submits that disclosure could enable transfer of the assets in dispute. What should the tribunal do?
A wallet is the software or device used to manage credentials needed to authorise transactions; it does not ordinarily contain the asset itself. A private key is secret cryptographic data used to generate a digital signature, while the corresponding public key allows that signature to be verified without revealing the private key. Unlike an ordinary document, however, a private key offered as evidence may remain capable of authorising a transaction.
Published arbitral decisions addressing private-key production appear scarce. The absence of settled practice matters because a tribunal faced with such a request risks either excluding potentially probative evidence of signing capability or creating a new security risk by ordering disclosure of a live credential. General arbitral powers permit inspection, experts, confidentiality measures and adverse inferences, but provide no credential-specific order of escalation. Since there is no evident settled practice, this post proposes a procedural model for how a tribunal might consider evidentiary issues related to the ability of a party to authorise a transaction.
This post proposes a five-rung verification ladder. Rungs one to four test the relevant proposition without production; rung five permits controlled production, and only in exceptional circumstances. Within rung five, access should progress from neutral-only inspection, to restricted inspection by a named representative of the opposing side, and, only extraordinarily, to direct access by specified party personnel. Unrestricted possession of an operational credential should not be ordered.
Why Ordinary Production Is the Wrong Starting Point
A private key can be copied without depriving the existing holder of the original, or necessarily revealing that a copy was made. A recipient of a usable copy may later generate valid signatures.
An undertaking not to use or copy the key is a legal obligation, not a technical guarantee. Once the complete credential has been exposed, the tribunal cannot reliably ensure that it was not retained. Disclosure may also weaken later attribution: if several people can now sign, a subsequent signature proves less about who exercised control.
Prior to any document disclosure order, the tribunal must therefore identify the precise fact to be proved. For example, current signing capability, historical access, account permissions, human identity, contractual authority, ownership and legal entitlement are different facts to be proved. Proof of one does not establish the others.
A Five-Rung Verification Ladder
The ladder below proposes to tailor disclosure obligations to the specific allegation. Its rungs increase in credential exposure, burden and security risk. The tribunal should use the lowest reliable rung and may skip an inapplicable one.
The requesting party should explain why a lower-risk method is insufficient; the responding party should identify the concrete risk and propose a workable alternative. Refusal of a proportionate method may affect evidentiary weight or support an adverse inference under the IBA Rules on the Taking of Evidence.
First, use public and corroborative records. A public blockchain may show that a transaction occurred, when it was recorded and which addresses were involved. Custody records, access logs and contemporaneous communications may connect that event to a person or organisation. These materials may establish the event and help link it to a person without exposing a live credential. They do not by themselves prove who used the credential, whether that person had authority, or who owned the asset.
Second, use a challenge-and-sign procedure. The tribunal or a neutral expert supplies a unique, one-off message. The party uses the relevant signing system to produce a digital signature, verified using the corresponding public key or a protocol-supported method. A valid result shows that signing capability was available for that message at that time without revealing the private key. It does not, without corroboration, identify the human user, establish historical control, or determine ownership. If current signing capability is the issue, there is ordinarily no reason to move to the next rung.
Third, use intermediary confirmation where the arrangement is custodial or otherwise intermediated. An exchange or custodian may manage the private keys while the user acts through an account and permissions; the user may therefore have no key capable of production. In such cases, the tribunal may direct a party, to the extent that this is within that party's legal power or control, to request or authorise a confirmation from the intermediary. The intermediary might confirm the account, balance, withdrawal permissions, approval thresholds or access records. This method tests the account structure, not possession of a self-custodied key or legal entitlement. If that confirmation resolves the account-control question, inspection of a credential is not warranted.
Fourth, use neutral expert observation of the configuration but not the key itself. An expert might watch a party unlock a hardware wallet (a physical signing device), confirm the displayed address, sign the tribunal's message and verify the result, all without seeing or retaining the private key. The expert’s mandate should define the device, the environment, the observed steps, and the record that must be preserved. The expert reports technical observations; the tribunal determines their legal effect. If observation resolves the issue, the credential remains out of production.
Fifth, consider controlled inspection of the key itself. This rung is justified only where the credential's content or identity—not merely the ability to use it—is materially in issue. For example, a party may need to show that an encrypted file or backup handed over under a custody agreement actually contains the correct key, or that a key component provided by one party is in fact the piece required under a multi-signature or key-splitting arrangement.
Controlled inspection may take three forms, considered in that order: neutral-only inspection, restricted access by named external counsel or a designated technical expert for the opposing side, and, only extraordinarily, direct access by specified party personnel. Opposing-party access is therefore an option, but it should not ordinarily mean delivery of a reusable copy or inclusion of the secret in the case file.
Before allowing the opposing-side access, the tribunal should find that lower rungs cannot provide reliable evidence and neutral-only inspection would not permit meaningful adversarial testing. It should identify everything the key can access or authorise, record the position before inspection begins, and—where technically possible—ensure the key can no longer be used to carry out transactions, for example by replacing or disabling the key, changing access rights, or moving the affected assets to a different address. An undertaking not to copy a live key is a legal safeguard, not a technical substitute for neutralising its operational effect.
The inspection should occur in an offline or otherwise isolated environment, with named attendees and an audit record excluding the secret. The tribunal's order should define what may be viewed, tested and recorded, how temporary material will be sealed or destroyed, and the response to a potential security incident.
Conclusion
The answer to the question in the title of this post is not "never", but "only exceptionally". Rungs one to four provide verification without production. The fifth becomes available only where the credential itself is materially in issue, lower-risk methods cannot provide reliable proof, and fair testing cannot otherwise be achieved.
Even then, production should mean controlled access under a security protocol, not unrestricted disclosure. The starting point should be inspection by a neutral expert alone; only where that proves insufficient should the tribunal consider restricted access by a named representative of the opposing side, with direct party access reserved for the rarest case and operational effect neutralised wherever possible. Production should be the fifth rung, not the tribunal's first instinct.
You may also like