THE AUTHOR:
Ali Shah Nawaz Niazi, Legal Trainee at Freshfields
Introduction
An award arrives. It is detailed, carefully reasoned, and signed by the arbitrator the parties selected. Nothing on its face suggests that anyone else played a meaningful role in deciding the dispute.
But unknown to the parties, behind the award sits an associate at the arbitrator’s firm. The associate has read the pleadings, reviewed the evidence, researched disputed legal issues, and prepared substantial parts of the reasoning. The arbitrator has reviewed the draft, made some changes, and signed it.
Is this an abdication by the arbitrator? Is this an improper delegation of the decision-making duty of the arbitrator? Was the arbitrator assisted, or was part of the adjudicative mandate delegated? And even more importantly, how would the parties ever know about this ‘behind-the-scenes’ activity and address it?
This is the real difficulty with tribunal secretaries and other arbitral assistants. The problem is not that arbitrators receive help. The problem starts when substantive assistance is invisible to the parties, when an approved role quietly expands, or when assistance crosses into the judgment entrusted to the arbitrator personally.
This is not simply a question of drawing the legal line between assistance and delegation. It is also a question of how this ‘behind-the-scenes’ activity can be verified because, unless it is detected, it cannot be actioned other than to be discouraged beforehand. The evidence capable of showing what actually happened, e.g. drafts, instructions, metadata, time records and internal communications, will usually sit inside the tribunal’s confidential working process. A party may therefore carry the burden of proving improper delegation while having almost no access to the material needed to do so.
Yet the obvious solution — giving a challenging party access to the tribunal’s drafts, instructions, time records and internal communications — creates another problem. If such access followed merely from an allegation that an assistant influenced the award, almost every significant award could generate a fishing expedition. The cure would be as damaging as the disease.
Arbitral institutions therefore need a middle course: enough transparency to make concealed delegation harder to hide and capable of limited verification, but a sufficiently high threshold to protect tribunals from speculative post-award challenges.
Where Does the Assistance End?
The starting point is not that drafting becomes permissible once disclosed. Disclosure is only one safeguard: the task must be authorised and supervised, and the tribunal must itself reach the conclusions expressed in the award.
In Total Support Management (Pty) Ltd v. Diversified Health Systems (SA) (Pty) Ltd [2002] ZASCA 14, the South African Supreme Court of Appeal accepted that an assistant could conduct research, discuss the merits and even prepare a first draft of an award. The decisive issue was whether the arbitrator had exercised his own judgment. That could be tested because the arbitrator explained his instructions, prior views and subsequent revision of the draft, supported by his fee records; no contrary facts were produced. The Yukos set-aside proceedings point in the same direction. Evidence concerning the tribunal assistant’s unusually extensive hours and involvement in drafting did not, by itself, establish that the arbitrators had delegated their mandate.
That distinction matters. A categorical ban on drafting would confuse authorship with decision-making. An award may be drafted by another hand while still reflecting the arbitrator’s conclusions. Conversely, an arbitrator may personally type every word while relying on someone else to perform the intellectual work that matters.
The test should therefore be functional rather than authorship-based. Who evaluated the evidence? Who resolved disputed facts and law? Who chose between competing experts? Who determined liability, quantum and relief? And whose judgment does the award ultimately express?
Research, chronologies, summaries, administrative coordination and controlled drafting can all remain forms of assistance. What cannot be transferred is the tribunal’s responsibility for deciding the issues in dispute.
Institutional rules increasingly reflect this distinction, although they do so in different ways.
Article 19 of the Abu Dhabi International Arbitration Centre (“arbitrateAD”) Arbitration Rules 2024 contains a relatively developed regime. A tribunal secretary may be appointed only with party approval; the parties are consulted on the secretary’s tasks, the secretary must provide a statement concerning availability, impartiality and independence, and the tribunal may not delegate decision-making authority.
Dubai International Arbitration Centre (“DIAC”) takes a different approach. The DIAC Arbitration Rules 2022 contain no dedicated provision governing tribunal secretaries or other persons assisting the tribunal. The Rules confer broad procedural discretion, but they do not create an express regime requiring disclosure of a tribunal assistant, defining the assistant’s role or prohibiting the delegation of adjudicative functions.
The ICC Rules provide a useful comparative benchmark. Article 44 of the ICC Rules 2026 permits appointment of a tribunal secretary after consultation with the parties, prohibits delegation of decision-making authority and subjects the secretary to independence, impartiality and confidentiality requirements.
Under Article 14A of the LCIA Rules 2020, the tribunal may obtain assistance from a tribunal secretary only after the parties have approved the secretary and the tasks to be performed, although agreement may be deemed where a party does not object within the period fixed by the tribunal. If additional tasks are later proposed, Article 14.11 requires the parties’ prior agreement. The LCIA Guidance Note for Parties and Arbitrators goes further. Section 10, paras. 201–202 contemplate that, with the parties’ approval, a tribunal secretary may attend deliberations, review authorities, and even prepare first drafts of awards or sections of awards.
These approaches differ in detail, but they share a common feature: they are principally preventive. They regulate who may assist, what that person may do, and what must be disclosed. The harder problem begins when those safeguards are breached out of sight.
The Party Outside the Room
The cases show how unusual it can be for a party to obtain meaningful evidence of what transpired behind the scenes in the tribunal’s award-making process.
In P v. Q [2017] EWHC 148 (Comm) and [2017] EWHC 194 (Comm), suspicion arose because the chairman mistakenly sent to a party’s lawyers an email intended for the tribunal secretary, asking for the secretary’s reaction to a procedural development. The challenging party then relied on that accidental email and time records to argue that the secretary had become too involved. Such a slip will of course occur only very rarely.
In Sonatrach v. Statoil [2014] EWHC 875 (Comm), the parties had agreed that the tribunal secretary would perform only administrative functions and would not participate in the decision. The challenging party later alleged that the secretary had prepared substantive notes for the tribunal. But when those notes were sought, the tribunal refused production on the basis that they formed part of the confidentiality of its deliberations. This shows the extensive difficulty in determining what occurred in the award-making process behind the scenes.
The tension is obvious. Confidential deliberations protect independent decision-making. But the same confidentiality can make it exceptionally difficult to determine whether that decision-making remained confined to the people authorised to participate in it.
There is also an evidentiary asymmetry. Where a secretary is disclosed, the parties know who is involved and what tasks were authorised, so any later departure can at least be identified against that agreed role. An undisclosed associate leaves no such reference point.
That is the verification gap.
Closing the Gap Without Opening the Files
Because post-award proof is so difficult, the safeguards must begin during the arbitration.
The first reform should be definitional: institutional rules should regulate assistance by function rather than title. A rule dealing only with a formally appointed “tribunal secretary” is too narrow. The relevant person may be called an associate, legal assistant, junior lawyer, employee, or something else entirely. If that person receives substantive case material or assists with factual, legal, evidentiary, or drafting work, the same safeguards should apply.
A useful umbrella term would be “Tribunal Support Person”.
Second, disclosure should occur before a Tribunal Support Person receives substantive case material. The parties should be told who the person is, the person’s professional relationship with the arbitrator, the proposed functions and any circumstances relevant to independence or impartiality. If the role later expands materially, for example, from administration to legal research or substantive drafting, the parties should be told again.
That improves transparency, but disclosure alone does not solve the harder problem: a rule can still be ignored. That is where institutional design can do something that ordinary disclosure rules cannot. Institutions could require the tribunal to lodge a short confidential ‘support register’ with the institution. It would record only each Tribunal Support Person’s identity, period of involvement and broad categories of work. It would not record the tribunal’s views on the merits, conclusions or draft reasoning. It would create an audit trail without creating a file of arbitral deliberations.
The final award could then contain a simple certification: all persons who provided substantive tribunal assistance were disclosed, and no adjudicative function was delegated. The mechanism would not be foolproof, but the obligation to create and certify a contemporaneous record should deter concealment and reduce the risk of undisclosed delegation.
The register should remain confidential and should not be disclosed merely because a party challenges the award. Access should arise only if the party first meets the evidentiary threshold discussed below.
A Gate Before the Door
A disappointed party should not be entitled to investigate the tribunal simply because the award is long, sophisticated, stylistically different from earlier procedural orders, or known to have involved drafting assistance.
Before any further review, the applicant should have to produce specific and objectively credible facts giving rise to a reasonable inference of improper delegation.
A misdirected email may qualify. Metadata identifying an undisclosed author may qualify. An admission, unexplained time records or other concrete evidence linking an undisclosed person to substantive decision-making may qualify.
Suspicion alone should not.
If that threshold is crossed, review should proceed in stages. The institution holding the register, or the supervisory court where appropriate, could first inspect it to compare the disclosed role with the allegation made. If the issue remains unresolved, limited metadata, task descriptions or time records could be reviewed under an appropriately restricted confidentiality arrangement. Drafts or internal communications should remain a last resort, in camera, and only to the extent necessary to answer the delegation question.
The challenging party needs a reliable answer. It does not need a general right to rummage through arbitral deliberations.
Not Every Breach Should Imperil the Award
The remedy must fit within the existing setting-aside and enforcement framework. In Model Law jurisdictions, improper delegation may engage Article 34(2)(a)(iv), where the tribunal’s composition or arbitral procedure was not in accordance with the parties’ agreement or applicable law. At the enforcement stage, Article V(1)(d) of the New York Convention provides the corresponding ground.
Failure to disclose a person who performed only permissible support work should not automatically imperil an award. Nor should the fact that a disclosed secretary exceeded an agreed role necessarily establish that the tribunal surrendered its adjudicative mandate.
The more serious case is where an unauthorised person materially decides disputed issues of fact or law, evaluates decisive evidence, resolves liability or quantum, or otherwise exercises judgment entrusted to the appointed tribunal. The complaint is then no longer merely that an assistant was used, but that the agreed arbitral process — or potentially the decision-making tribunal itself — was not the one the parties selected.
At the post-award stage, the corresponding reform should therefore be procedural: once an applicant meets the prima facie threshold, the competent court should be empowered to use narrowly protected means of verifying the allegation.
That is the balance.
Conclusion
Arbitration asks parties to accept that much of the tribunal’s work will happen behind a closed door. That is both inevitable and, in many respects, desirable. But the legitimacy of that closed process still depends on one basic proposition: the people appointed to decide the dispute must actually be the people who decide it.
Institutions do not need to open the door completely. They need a way to check, in the exceptional case where there is real evidence of a problem, who was behind it.
Assistance may be shared. The adjudicative mandate cannot.
