By Ibrahim Sorie Yillah *
Abstract
On 24 July 2026 the Assembly of States Parties (ASP) to the Rome Statute voted to remove Karim A A Khan KC from office as Prosecutor of the International Criminal Court, notwithstanding a unanimous finding by an ad hoc panel of three senior judges that the facts established by the Court’s own oversight body did not amount to misconduct. This article argues that the process by which that removal was reached was legally defective in several distinct respects: it re-characterised the factual record beyond what the judges had found; it discarded a judicial recommendation without a reasoned legal justification; and it substituted, for the deliberative rigour ordinarily associated with disciplinary adjudication, a secret ballot of twenty-one and then one hundred and twenty-five state representatives. The article asks bluntly whether anything resembling a trial occurred at all, traces the asymmetry between Article 46’s treatment of judges and its treatment of the Prosecutor, and considers what the episode portends for the independence the Rome Statute was designed to guarantee to future holders of the office of the Prosecutor.
I. Introduction
Institutions that judge others are, from time to time, called upon to judge themselves. On 24 July 2026, at a special session convened at United Nations Headquarters in New York, the Assembly of States Parties to the Rome Statute did exactly that: by voting to remove Karim Khan from office as Prosecutor of the International Criminal Court, on findings that he had committed serious misconduct and a serious breach of duty.[1] The vote closed eighteen months of inquiries that had, by the end, drawn public criticism from more than one hundred and eighty civil society organisations, a formal legal opinion from senior English counsel warning that the process risked unlawfulness, and open dissent within the Bureau of the Assembly itself.
The removal did not occur in a political vacuum. It came amid explicit threats from the government of the United States to dismantle the Court “brick by brick,” amid sanctions imposed on a substantial number of the Court’s judges and prosecutorial staff, and in the long shadow of Mr Khan’s own decision, two years earlier, to seek arrest warrants against the Prime Minister of Israel, his former Defence Minister, and three leaders of Hamas.[2] That context does not by itself prove that the removal was politically motivated. But it does mean the process by which it was reached must be scrutinised with particular care, precisely because an institution whose central promise is impartial adjudication cannot afford to have its own governance appear to bend to the political weather.
This article advances three connected claims. First, that the procedure culminating in the Prosecutor’s removal departed, in material respects, from the requirements of Article 46 of the Rome Statute and from ordinary standards of administrative due process, such that the removal is properly characterised as legally defective. Second, that the manner of the departure — a political organ overriding a unanimous judicial finding without publishing a reasoned legal basis for doing so — threatens the independence mandate the Statute was designed to secure for the Office of the Prosecutor, with consequences reaching well beyond Mr Khan personally to every future holder of the office. Third, and most pointedly, that the process never amounted to a trial in any sense recognisable to administrative or criminal law, and that the Assembly’s own governing body allowed twenty-one ambassadors, and then one hundred and twenty-five, to usurp a function the Court’s internal architecture had assigned to judicial expertise. The article takes no position on whether Mr Khan did or did not engage in the conduct alleged; that question was, after lengthy investigation, unanimously resolved by the very judges appointed to answer it. The argument here concerns process, not the underlying facts, and it is on process that institutions asking the world to trust their fairness are most exposed when they fall short. To sharpen that argument, the article also compares two national impeachment trials — of Kenya’s Deputy President in 2024 and of the President of the United States in 2020 — in which the political body empowered to remove a senior officer conducted, at least formally, a trial with live evidence and cross-examination before voting; the comparison is used not to suggest such a hearing was compelled by the Rome Statute, but to expose how far short of ordinary comparative practice the ASP’s own procedure fell.
II. The Architecture of Article 46
Article 46 of the Rome Statute is the Court’s own dismissal clause. It permits removal of a judge, the Prosecutor, a Deputy Prosecutor, the Registrar or a Deputy Registrar on three grounds: serious misconduct or a serious breach of duty under the Statute or the Rules of Procedure and Evidence, or an inability to exercise the functions required.[3] What matters greatly to the argument developed here is that the Statute does not treat all of these office-holders alike once removal is contemplated.
For a judge, Article 46(2)(a) requires that removal be decided by the Assembly by a two-thirds majority of States Parties, but only after the other judges of the Court have themselves recommended removal by a two-thirds majority.[4] That is a double lock: a professional, judicial screen sits in front of the political vote, and unless that screen produces a positive recommendation, the Assembly’s hands are tied. The screen exists precisely because judges must be shielded from the momentary political preferences of the states whose conduct they may one day be asked to judge.
For the Prosecutor, by contrast, Article 46(2)(b) requires only an absolute majority of States Parties. There is no equivalent requirement that a body of judicial peers first recommend removal, and the Statute is, on its face, silent as to how the Assembly is to satisfy itself that serious misconduct has occurred.[5] The asymmetry is not an oversight; it reflects the different constitutional position of the Prosecutor, who is an executive rather than an adjudicative officer of the Court. But the absence of a textual double lock makes the procedural safeguards that do exist — chiefly the guarantee in Article 46(4) that the officer under scrutiny shall have “full opportunity to present and receive evidence and to make submissions” — carry correspondingly greater weight. Rule 27(2) of the Rules of Procedure and Evidence echoes the same guarantee.[6] Where the political check is thin, the procedural check must be robust. It is against that yardstick that the events of 2025 and 2026 are properly measured.
III. The Ad Hoc Panel: Genesis, Mandate and Findings
The mechanism that came to occupy the centre of this controversy — an ad hoc panel of independent judicial experts — traces to a recommendation of the 2020 Independent Expert Review of the Court, which the Assembly partially endorsed and later formalised in Resolution ICC-ASP/24/Res.2 as part of the Court’s disciplinary architecture.[7] Its evident purpose was to inject legal and judicial rigour into what would otherwise be a purely administrative assessment, by diplomats untrained in evaluating evidence, of investigative findings.
In November 2024, after the complainant against Mr Khan declined to engage the Court’s own internal investigative mechanism, the President of the Assembly, Paivi Kaukoranta, authorised referral of the matter to the United Nations Office of Internal Oversight Services (OIOS) for independent fact-finding.[8] A caveat is necessary here: the OIOS report has never been made public, nor has the Bureau’s internal record of its own deliberations. What follows accordingly draws on press reporting, on public statements by Mr Khan’s counsel, and on a coalition statement issued by civil society organisations — sources themselves secondary to, and in places interpretive of, a confidential process. Figures such as the length of the OIOS report or the number of its findings are reported rather than independently verified. According to press accounts, OIOS conducted an extensive investigation, interviewing witnesses and reviewing electronic communications over many months.[9] One commentator has suggested the investigation generated some one hundred and thirty-seven discrete findings, none characterised by OIOS itself as amounting to misconduct; that figure cannot presently be checked against the report and is recorded here only as an unverified claim.[10] What can be stated with greater confidence, resting on the ASP President’s own subsequent public statement, is that the OIOS findings were referred to a three-member panel of senior judges: Justice Leona Theron of South Africa’s Constitutional Court, Judge Paul Lemmens of Belgium, and Sir Seymour Panton of Jamaica.
The panel spent nearly three months examining the OIOS material. On 9 March 2026 it reported, unanimously, that the facts established by OIOS did not amount to misconduct or a breach of duty under the Court’s own legal framework.[11] That finding was not a technicality. It was the considered, unanimous view of three senior sitting and former judges, applying the same evidentiary standard the Court applies in its ordinary work, after months of careful review of the entire investigative record.
IV. From Recommendation to Rejection: The Bureau’s Pivot
What followed is the fulcrum on which this article’s illegality argument turns. Within weeks, a majority of the twenty-one-member Bureau of the Assembly voted to set the panel’s findings aside, reaching what it called a preliminary determination that the material warranted further examination.[12] Two months later the Bureau formally suspended Mr Khan and referred the matter to the full Assembly, having concluded by majority that he had committed serious misconduct.
Before the Bureau took that step, it had been warned in the plainest terms that it was heading toward unlawful ground. Sir James Eadie KC, instructed by Mr Khan’s legal team, submitted a twenty-one-page opinion criticising the “restrictive mandate” the Bureau itself had given the three-judge panel. His central point deserves to be stated plainly: by confining the panel to assessing whether OIOS’s own findings amounted to misconduct, while denying it any independent fact-finding or adjudicative power, the Bureau had engineered a process with no judicial or independent fact-finding phase at all — a design he warned would be unlawful and unsustainable in principle.[13] The Bureau proceeded regardless.
A further caveat applies here. Neither the Bureau’s internal record nor the OIOS report has been published, and the account below is drawn from a statement issued by a coalition of more than one hundred and eighty Palestinian and international civil society organisations, itself reporting on a process it did not directly observe. That coalition’s account — treated here as reported and contested rather than established — is that the Bureau majority, comprising among others Belgium, Brazil, Chile, Finland, Italy, Japan, New Zealand, Poland and Switzerland, did not adopt the panel’s unanimous legal conclusion, and instead reached a finding, in the coalition’s own words, that Mr Khan had “engaged in a sexual relationship” with the complainant, and that such a relationship could never be appropriate given the power imbalance between them.[14] On the same account, neither Mr Khan nor the complainant had ever described their contact in those terms, and this particular characterisation does not appear, so far as the coalition could establish, in the OIOS report or the panel’s own findings — though neither document is public, and the point cannot presently be verified.[15] If accurately reported, this divergence between the Bureau’s stated basis for removal and the record reviewed by OIOS and the panel would matter greatly to the due process argument developed below. Sierra Leone was reported to be among the states that voted against the Bureau’s position, alongside South Africa, Kenya and Senegal.[16]
V. Was There a Trial? Characterising the Process
It is worth asking the question directly, because the language used throughout — finding, determination, serious misconduct — borrows heavily from the vocabulary of adjudication, and borrowed vocabulary can obscure an absence of adjudicative substance. The ASP President’s own explanatory statement, issued three days after the vote, is unusually candid on this point. It confirms that the panel’s mandate was advisory, that the Bureau retained responsibility for making determinations on the allegations, and that the panel’s report did not exonerate Mr Khan only because its mandate did not permit additional investigative steps.[17] Put plainly: the Bureau designed a panel with no power to find facts of its own, asked it only whether OIOS’s facts supported a legal conclusion of misconduct, received the answer no, and then treated the panel’s inability to go further as licence for the Bureau itself to reach a different factual conclusion — one involving a characterisation of the relationship that no investigator had made — without conducting any fact-finding exercise of its own resembling a hearing.
There was no adversarial hearing before the body that ultimately decided Mr Khan’s fate. There was no application, by that body, of a stated evidentiary standard. There was no cross-examination, no disclosure regime, and — most tellingly — no published, reasoned opinion explaining why three judges applying the Court’s own legal framework were wrong. What there was, instead, was a vote: first twenty-one ambassadors in the Bureau, then one hundred and twenty-five state representatives in the Assembly, voting by secret ballot on a resolution that used the vocabulary of judicial fact-finding without any of its structure. That is not a trial. It is, at most, a political ratification dressed in judicial language — and the distinction matters because the Rome Statute’s own drafters, when they wished to entrust removal decisions to judicial rigour, said so explicitly, as they did for judges under Article 46(2)(a).
VI. Learning from Comparative Impeachment Practice: Kenya and the United States
A comparison worth making explicit before the illegality thesis is developed further. In mature constitutional systems that vest the power to remove a senior public officer in a political body, that body does not typically content itself with voting on a paper record assembled by others. It sits, at least formally, as a trial forum: it hears a case presented by a designated prosecuting party, permits the officer under scrutiny to test that case through cross-examination, and only then votes. Two recent, well-documented examples illustrate the point.
A. The Kenyan model. Under Article 145 of the Constitution of Kenya, removal of the Deputy President begins with a resolution of the National Assembly and proceeds, if approved, to a trial before the Senate.[18] In October 2024 that is precisely what occurred: after the National Assembly approved eleven charges against then Deputy President Rigathi Gachagua by a vote of 282 members, the Senate received the charges and the documentary and electronic evidence formally laid before it, heard Mr Gachagua enter a plea of not guilty, and conducted what was explicitly styled a trial. The National Assembly’s legal team called live witnesses — among them the mover of the motion, a former state agency chief executive, the secretary to the Cabinet, the Governor of Nairobi and a private businessman — each subject to cross-examination by Mr Gachagua’s own counsel, while Mr Gachagua himself stood as the sole witness in his own defence.[19] Two full days were set aside for that cross-examination before the Senate voted, by the constitutionally required two-thirds majority, to remove him.[20]
The Kenyan process was not without its own difficulties, and these are instructive in their own right. When Mr Gachagua was hospitalised during the hearing on 17 October 2024, the Senate declined his counsel’s request for an adjournment and concluded the hearing in his absence.[21] A subsequent judicial review held that the process had, on the whole, been fair, but that this episode had infringed his fair trial rights — and, importantly, that the rights to fair administrative action and to a fair hearing under the Kenyan Constitution applied in full to a state officer facing removal, notwithstanding the avowedly political character of impeachment itself.[22] Even a legislative body exercising a constitutionally political function does not escape the ordinary discipline of natural justice merely because it is a legislature rather than a court.
B. The American model. The United States Constitution vests the Senate with “the sole Power to try all Impeachments,” and requires the Chief Justice to preside when the President himself is tried.[23] The House appoints managers who function as prosecutors at the bar of the Senate; the respondent is entitled to his own counsel; and the trial proceeds with opening arguments from both sides. In President Trump’s first Senate trial, in January 2020, the Senate allotted twenty-four hours over three days to each side for opening arguments alone, and separately debated — and in that instance rejected, largely along party lines — a proposal to subpoena live witness testimony.[24] That rejection is itself informative: even in a system with an explicit constitutional trial mandate, whether to hear live witnesses is a further question left to the discretion of the deciding body.[25] Indeed, neither of President Trump’s two Senate trials featured live witness testimony, and the 1999 Clinton trial permitted only pre-recorded depositions from three witnesses.[26] What remained constant across all these trials was the basic adversarial architecture: a designated prosecuting party presenting a case directly to the very body that would vote, with the respondent’s counsel permitted to answer before the same audience, followed by a two-thirds vote.[27]
C. Is a comparable hearing required, or precluded, by the Rome Statute? Nothing in the Rome Statute forbids the Assembly from adopting a hearing procedure of the kind used in Kenya or the United States. Unlike the American and Kenyan constitutions, which expressly assign a trial function to the deciding body, the Rome Statute is comparatively spare: Article 46 assigns the removal decision to the Assembly but is silent on how the Assembly is to satisfy itself of the underlying facts, while Article 46(4) guarantees only a “full opportunity to present and receive evidence and to make submissions” — language broad enough to accommodate a full adversarial hearing, but not language that compels one, and equally capable, on a minimalist reading, of being satisfied by an exchange of written submissions to the Bureau alone.[28] The Assembly’s own creation of the ad hoc panel in 2025 is itself proof that it possesses the institutional competence to design more elaborate quasi-judicial procedures should it choose to.[29] The absence of a Kenyan- or American-style hearing to date is therefore a matter of institutional choice, not statutory constraint.
Where the comparison bites hardest is not in showing that a trial was legally compelled, but in showing what the ASP’s actual procedure lacked relative to both comparators. In Kenya, the body that ultimately voted was the same body that received the evidence, watched the cross-examination, and formed its own view of the witnesses’ credibility, with a reviewing court on hand to correct the one significant departure from fair process. In the United States, even in the more minimalist of the two Trump trials, managers and defence counsel still presented their arguments directly to the assembled body that would vote, before a presiding judicial officer empowered to rule on evidence. In neither system did the deciding body outsource its fact-finding entirely to an external panel, receive that panel’s unanimous and reasoned conclusion, and then vote to reach a materially different account of the facts without itself hearing evidence on the point or explaining why the panel had erred.
The ASP procedure ended up with neither of the two forms of protection other systems supply. It did not have the double-lock judicial screen Article 46(2)(a) affords to judges, under which a positive judicial recommendation is a precondition to any political vote at all.[30] Nor did it have the direct evidentiary trial Kenya and the United States use for their own senior officers. What it had instead was an advisory judicial panel whose unanimous, reasoned conclusion could simply be set aside by the very body that had appointed it, without a hearing of its own and without a reasoned decision a reviewing body could test. Measured against either comparator, the ASP’s process sits below the floor both jurisdictions already treat as a minimum incident of due process — not above it, and not merely level with it. A structured hearing before the Assembly, or a delegated committee of it, would not be contrary to the Rome Statute; on this analysis it would bring the Court’s own accountability mechanism into closer alignment with practice that constitutional democracies as different as Kenya and the United States already regard as ordinary.
VII. The Illegality Thesis
A. Ultra vires re-characterisation of the conduct. This ground is necessarily framed conditionally, given that neither the OIOS report nor the Bureau’s internal record is public, and the account relied upon comes from a civil society coalition rather than the Bureau itself. On that reported account, however, a decision-making body that substitutes its own account of the underlying facts for the account given by investigators and confirmed — or here not confirmed — by the reviewing judges, without conducting any fact-finding process capable of producing that new account, acts beyond the function assigned to it. The Bureau was not constituted, and did not purport to sit, as a fact-finding tribunal. If the coalition’s account of a finding that a relationship existed is accurate, and unsupported by either party’s own account or by the OIOS record as reported, it would be difficult to characterise that finding as anything other than an ultra vires exercise of a function the Bureau’s own procedural design had withheld from itself. This ground is accordingly the most provisional of the five set out here, turning on facts only publication of the underlying record could settle.
B. Breach of the Article 46(4) due process guarantee. Article 46(4) and Rule 27(2) guarantee the officer under scrutiny a full opportunity to present and receive evidence and to make submissions.[31] That guarantee, properly read, must extend to the plenary body that ultimately decides the officer’s fate, and not merely to the twenty-one-member Bureau; the suspension power under Rule 28 is an interim measure directed at the active exercise of official duties, and cannot, without collapsing the distinction between interim measures and final deprivation of office, substitute for the right to be heard by the decision-maker itself.[32] On the public record, Mr Khan’s opportunity to answer the case before the Assembly that actually voted on his removal was considerably thinner than his opportunity to answer it before the panel whose findings the Assembly then discarded.
C. Departure from the Rules without transparent amendment. Reporting closer to the vote suggested the Bureau altered the sequencing and threshold of its own procedure part-way through, restructuring what had been billed as a two-stage process — a misconduct finding, followed by a separate removal vote — in a manner that lowered the practical threshold for removal.[33] Whatever the precise mechanics, a body conducting disciplinary proceedings against its own chief prosecutor cannot, consistently with basic administrative law principles, change the rules of the contest while it is under way without at minimum a transparent and reasoned explanation to the person affected.
D. Discarding a unanimous judicial finding without a reasoned countervailing basis. Perhaps the single most striking feature of this episode is one that would trouble any administrative lawyer regardless of subject matter: a political body rejected the unanimous legal conclusion of three senior judges, appointed by that same body for the express purpose of bringing judicial rigour to the assessment, without ever publishing a reasoned opinion identifying where those judges had erred.[34] A reviewing body which disagrees with a judicial panel it created ordinarily owes that panel, and the person affected, an explanation grounded in law.
E. Politicisation and extraneous considerations. None of this occurred in isolation from the wider geopolitical storm around the Court. The proceedings unfolded against a backdrop of United States sanctions on a substantial number of the Court’s judges and staff, explicit threats to dismantle the institution, and continuing controversy over the arrest warrants Mr Khan’s office had sought against the Israeli Prime Minister and his former Defence Minister.[35] That context does not establish, and this article does not assert, that any individual state’s vote was cast in bad faith. But where a process already carries the procedural infirmities catalogued above, the surrounding political pressure is a legitimate interpretive aid in assessing whether the departure from ordinary due process was accidental or expedient.
VIII. Ambassadors as Adjudicators: A Separation-of-Powers Problem
The deeper structural point is one of institutional design. The Rome Statute’s drafters, confronted with the question of who should judge a judge, answered it by requiring the judiciary to judge itself first, subject only then to political ratification by a supermajority of states.[36] When it came to the Prosecutor, the drafters left a gap, and the Assembly’s own creation of the ad hoc panel mechanism in 2025 was a sensible attempt to fill that gap with an analogous, if advisory, judicial screen. The events of 2026 demonstrate the danger of leaving such a screen merely advisory: once its conclusion proved politically inconvenient, the very body that had created the screen simply stepped around it.
A useful, if inexact, analogy may be drawn from company law. A board of directors that commissions its own audit committee to investigate an allegation against the chief executive, receives a unanimous finding that the allegation is not made out, and then dismisses the chief executive anyway on a materially different characterisation of the facts the audit committee never reached, would not ordinarily survive judicial review in most common law jurisdictions without, at the very least, a fresh and properly constituted inquiry into the new allegation. International institutions asking the world’s domestic courts, and the world’s citizens, to trust their internal governance can reasonably be held to a standard no lower than company law would impose on a private board.
Twenty-one ambassadors sitting as the Bureau, and one hundred and twenty-five state representatives sitting as the Assembly, are diplomats performing a diplomatic and political function. That is entirely proper for the great many governance tasks the Assembly performs — budgets, elections, treaty amendments, cooperation resolutions. It is markedly less well suited to the specific task of finding, as a matter of fact and law, that a named individual engaged in serious personal misconduct, particularly once the Assembly’s own chosen mechanism for injecting judicial competence into that exercise had already spoken, and had spoken the other way.
IX. Implications for the Independence Mandate of Future Prosecutors
Article 42 of the Rome Statute requires the Office of the Prosecutor to act independently as a separate organ of the Court and prohibits its members from seeking or acting on instructions from any external source.[37] Security of tenure — the assurance that a Prosecutor cannot be removed except through a process that is itself scrupulously fair — is not incidental to that independence; it is one of its principal structural guarantees. A Prosecutor who can be removed notwithstanding a unanimous judicial finding in his favour, through a process a former United Nations oversight official and senior counsel both warned risked unlawfulness, has, in practical terms, less security of tenure than the text of Article 42 appears to promise.
The consequence is unlikely to be confined to Mr Khan. Every future Prosecutor who contemplates opening an investigation or seeking a warrant against nationals of a powerful state, or against the allies of such a state, will now have concrete recent evidence that a sufficiently determined and sufficiently large coalition of states can, in the last resort, engineer removal notwithstanding the findings of the Court’s own judicial safeguards. That is precisely the chilling effect Article 42 was designed to prevent. An institution that already depends on states for funding, cooperation and enforcement of its warrants — dependencies that have visibly frayed in recent years — cannot afford to add a further, self-inflicted incentive for its chief prosecutor to calibrate prosecutorial decisions to the tolerance of the powerful.[38]
X. A Fair Hearing of the Contrary View
It would be one-sided to leave the argument there without acknowledging its most serious challenge. A number of respected commentators, including some with long institutional ties to the Court, have argued that the Assembly is not an appellate court, that Article 46 vests final decisional authority in the Assembly and not in any advisory panel, and that alleged procedural defects should not be treated as dispositive of the ultimate removal question.[39] These authors point out, fairly, that the allegations were investigated for well over a year, that Mr Khan and his lawyers were given repeated opportunities to make submissions to the Bureau, and that the vote itself tracked the absolute-majority threshold Article 46(2)(b) actually requires.[40]
Those views deserve to be taken seriously, and this article does not dismiss them. But they do not, in the end, answer the narrower and more exacting due process objection pressed here. That the Assembly possesses final decisional authority under Article 46(2)(b) is not in dispute; the question is whether the exercise of that authority, on this occasion, was accompanied by the fact-finding rigour and reasoned explanation due process required when a political body departs from the unanimous conclusion of the judicial mechanism it created for precisely that purpose. Final authority to decide is not the same thing as unconstrained authority to decide however one pleases; even bodies that sit at the apex of a decision-making hierarchy remain bound by the ordinary requirements of fair process, particularly where, as here, the office-holder’s independence from political pressure is itself a value the same instrument was written to protect.
XI. Recommendations
Several modest reforms would go a considerable distance toward closing the gap this episode has exposed, without requiring wholesale amendment of the Rome Statute.
First, the mandate of the ad hoc judicial panel should be placed on a firmer and more clearly binding footing, such that its unanimous conclusions may be displaced by the Bureau or Assembly only through a published, reasoned opinion identifying a specific legal or evidentiary error in the panel’s reasoning — mirroring, in substance if not in form, the double-lock protection Article 46(2)(a) already affords to judges.
Second, the right to be heard under Article 46(4) should be given practical content before the plenary decision-making body itself, and not treated as satisfied by submissions made only to the Bureau at an earlier stage.
Third, any change to the sequencing or voting thresholds of a disciplinary process already under way should require advance notice to the officer concerned and a stated legal justification.
Fourth, the Assembly should consider structurally separating the Bureau’s role in referring and managing a disciplinary complaint from its role, if any, in assessing the ultimate merits, so as to avoid the appearance of a single body acting as both investigator and judge.
Fifth, every removal decision under Article 46 should be accompanied by a published, reasoned decision setting out the factual and legal basis for the outcome, in language those trained in the analysis and evaluation of evidence in administrative and criminal law, and not only a diplomat, could test against the record.
Sixth, and drawing directly on the comparative practice examined in Part VI, the Assembly should consider adopting, whether by resolution or by amendment of the Rules, a structured hearing procedure — modelled on, though not necessarily identical to, the trial-style proceedings used in Kenya’s Senate and the United States Senate — under which a designated body within the Assembly hears the case against the officer directly, permits cross-examination of contested evidence, and reaches its own reasoned view, rather than voting on a record it never itself examined. Nothing in the Rome Statute precludes such a procedure, and the Assembly’s own establishment of the ad hoc judicial panel demonstrates that it already possesses the institutional means to create one.
XII. Conclusion
This article has not sought to adjudicate the truth of the allegations against Mr Khan; that task was undertaken, at considerable length, by investigators and by three senior judges applying the Court’s own legal framework, and their unanimous conclusion was that the facts established did not amount to misconduct. What this article has sought to show is that the process by which the Assembly nonetheless removed him departed, in several distinct and cumulative respects, from the ordinary requirements of administrative due process: it re-characterised the facts beyond anything the investigators or judges had found; it discarded a unanimous judicial recommendation without a reasoned legal explanation; and it relied, for its ultimate factual and legal conclusions, on a body of diplomats never designed, and institutionally ill-suited, to perform an adjudicative function of that kind.
The Rome Statute was built on the premise that even the gravest accusations deserve a fair process before an impartial tribunal. An institution that asks the world to accept that premise for the defendants who appear before it can reasonably be asked to extend no lesser a courtesy to its own chief prosecutor. Whatever else the removal of Karim Khan settles, it has not, on the record examined here, settled that question satisfactorily — and the independence the Statute promises to every future Prosecutor is measured only as strongly as the process by which the last one was made to leave.
*Ibrahim Sorie Yillah is A Member of the Sierra Leone Bar, A Former Trial Lawyer, Office of the Prosecutor, ICC, Former Defence Co-Counsel, ICC, Outgoing Member, Trust Fund For the Benefit of Victims, ICC, Principal Defender, Residual Special Court for Sierra Leone, A Tutor in International Criminal Law, Sierra Leone Law School. The views expressed in this article are solely those of the Author
[1]Assembly of States Parties, ‘Press Release: Assembly of States Parties Concludes its Special Session on Disciplinary Proceedings Involving an Elected Official’ (ICC, 24 July 2026); the Assembly voted 82 in favour, 13 against, with 15 abstentions.
[2]See eg ‘ICC Removes Karim Khan as Leadership Crisis Collides With US Pressure’ EU Today (26 July 2026); Janet H Anderson, ‘ICC: Karim Khan Removed – What Next?’ Justice Info (25 July 2026).
[3]Rome Statute of the International Criminal Court (adopted 17 July 1998, entered into force 1 July 2002) 2187 UNTS 3 (‘Rome Statute’), art 46(1)(a)-(c).
[4]Rome Statute (n 3) art 46(2)(a).
[5]Rome Statute (n 3) art 46(2)(b); see also ‘The Institutional Process against the Prosecutor: What States Must Do at the ASP (Part I)’ Opinio Juris (18 July 2026), observing that the Statute ‘is silent on how the assessment by the ASP of possible serious misconduct shall be made’ for the Prosecutor, in contrast to the two-thirds judicial recommendation required for a judge.
[6]Rome Statute (n 3) art 46(4); International Criminal Court, Rules of Procedure and Evidence, ICC-ASP/1/3 (2002, as amended) (‘RPE’), r 27(2).
[7]International Criminal Court, ‘Statement of the ASP President Following the Conclusion of Disciplinary Proceedings Involving Prosecutor Karim Khan’ (ICC-CPI, 27 July 2026), tracing the Panel mechanism to the 2020 Independent Expert Review (IER), partially endorsed by the Assembly, and formalised in Assembly of States Parties, Resolution ICC-ASP/24/Res.2 (2025).
[8]‘Former UN Oversight Chief Says ICC Prosecutor Karim Khan is “Victim of Massive Stitch-Up”‘ Middle East Eye (17 July 2026), noting that the ASP President authorised referral to OIOS in November 2024 after the complainant declined to engage the Court’s internal investigative mechanism.
[9]‘Karim Khan Ousted as ICC Prosecutor’ Justice for Justice (24 July 2026), reporting an OIOS report of some 150 pages drawing on over 5,000 pages of underlying material; the report itself has not been made public, and the figures are drawn from press reporting.
[10]Kenneth Roth, reported in ‘Karim Khan Dared to Hold Israel to Account – and That’s Why He Was Brought Down’ Middle East Eye (27 July 2026), citing a reported 137 discrete OIOS findings, none said to amount to misconduct; unverified against the undisclosed OIOS text.
[11]‘Karim Khan Ousted as ICC Prosecutor’ Justice for Justice (24 July 2026); the panel comprised Justice Leona Theron (South Africa), Judge Paul Lemmens (Belgium) and Sir Seymour Panton (Jamaica), reporting on 9 March 2026.
[12]‘ICC Member States Remove Prosecutor Karim Khan by Majority Vote’ Middle East Eye (24 July 2026), reporting that a Bureau majority voted, within weeks of the Panel’s report, to set the findings aside; the Bureau thereafter suspended Mr Khan and referred the matter to the Assembly.
[13]‘Exclusive: UK Government Lawyer Warned ICC Bureau its Khan Disciplinary Process “Unlawful”‘ Middle East Eye (7 July 2026); the opinion is attributed to Sir James Eadie KC, instructed by Mr Khan’s legal team.
[14]Over 180 Palestinian and International Organisations, ‘Lack of Due Process and Politicisation in Khan Disciplinary are Corroding the Independence of the ICC’ (Al-Haq, 14 July 2026), recording the Bureau vote by state and the coalition’s own account of the Bureau’s characterisation of the underlying conduct; the Bureau’s deliberations and the OIOS material remain confidential, so this account is reported and contested rather than verified.
[15]Over 180 Palestinian and International Organisations (n 14).
[16]Over 180 Palestinian and International Organisations (n 14).
[17]International Criminal Court, ‘Statement of the ASP President’ (n 7).
[18]Constitution of Kenya 2010, art 145; ‘Kenya: Court to Rule Tomorrow on Whether Senate Can Proceed With Gachagua Impeachment Trial’ AllAfrica (14 October 2024), recording that the National Assembly approved eleven charges against Deputy President Rigathi Gachagua by 282 votes before transmission to the Senate.
[19]‘Kenya: Gachagua to Be Sole Witness at Senate Trial, Sakaja, Mutuse Appear for National Assembly’ AllAfrica (16 October 2024).
[20]‘Kenya: Deputy President Gachagua Pleads Not Guilty in Senate Impeachment Trial’ Africanews (16 October 2024); removal required a two-thirds majority of the Senate.
[21]‘The Impeachment Process of His Excellency Gachagua Was Fair Save for Infringement of Fair Trial Process at the Senate Stage’ Kenya Law (case report, accessed 2026).
[22]Kenya Law (n 21), holding that the rights to fair administrative action under article 47 of the Constitution and to a fair hearing under article 50 fully accrued to a state officer facing removal, notwithstanding the political character of impeachment.
[23]US Constitution, art I, § 3, cl 6; Congressional Research Service, ‘The Impeachment Process in the Senate’ (Congress.gov, R46185).
[24]‘House Sends Impeachment of President Trump to Senate for Trial, Names 7 “Managers” to Prosecute Case’ USA Today (16 January 2020).
[25]‘Trump Impeachment: US Senate Approves Trial Rules’ Al Jazeera (22 January 2020), recording that a Senate motion to subpoena witness testimony from John Bolton was defeated largely along party lines.
[26]Congressional Research Service (n 23), noting depositions from three witnesses in the 1999 Clinton trial but no witness testimony on the Senate floor, and no witnesses examined in either Trump trial.
[27]Ballotpedia, ‘Impeachment of Federal Officials’.
[28]Rome Statute (n 3) art 46(4); RPE (n 6) r 27(2).
[29]International Criminal Court, ‘Statement of the ASP President’ (n 7).
[30]Rome Statute (n 3) art 46(2)(a).
[31]Rome Statute (n 3) art 46(4); RPE (n 6) r 27(2).
[32]Beatrice Bonafe and Ezequiel Jimenez Martinez, ‘Removing the ICC Prosecutor? A Roadmap for What Comes Next’ Opinio Juris (19 June 2026), arguing that the interim suspension power under RPE r 28 is directed at the Prosecutor’s active exercise of official duties and ‘does not and cannot extend to his basic right of defence’.
[33]‘Exclusive: ICC Bureau Changes Rules to Lower Threshold for Khan’s Removal’ Middle East Eye (27 June 2026).
[34]Kenneth Roth (n 10).
[35]See eg ‘ICC Removes Karim Khan as Leadership Crisis Collides With US Pressure’ (n 2); Anderson (n 2).
[36]Rome Statute (n 3) art 46(2)(a).
[37]Rome Statute (n 3) art 42(1) and (2).
[38]Anderson (n 2), noting instances of non-cooperation by States Parties in the enforcement of arrest warrants.
[39]Jennifer Trahan and Alex Whiting, ‘Before the Assembly: The Removal Vote and the Question of Fitness for Office’ Just Security (23 July 2026), arguing the Assembly ‘is not an appellate court’.
[40]Rebecca Hamilton and Alex Whiting, ‘The Future of the International Criminal Court: What Follows After Removal of the Prosecutor’ Just Security (24 July 2026).
