International Court of Justice

The International Court of Justice is the only principal organ of the United Nations that can issue a binding ruling against a sovereign state — and the only one that must wait to be asked. It sits at The Hague with fifteen judges and a Statute annexed to the Charter, and its docket is busier than at any point in its history.

It also has no police, no prosecutor and no power to open a case on its own initiative. Real legal authority resting entirely on consent is what makes the Court simultaneously the most law-like institution in the UN system and the one most easily dismissed as ornamental. What follows is how it came to be built that way, how it is composed, what it can hear, what its record shows, and what the enterprise says about whether international law does any work at all.

The road to a permanent world court

The Court did not appear in 1945 out of nothing. It is the third layer of a project that began with arbitration, hardened into a standing tribunal under the League, and was rebuilt with a new name at San Francisco.

Arbitration before adjudication

• Arbitration is settlement by a tribunal the parties themselves constitute, applying rules they choose; adjudication is settlement by a standing court with a fixed bench, procedure and applicable law. The move between them is the Court’s prehistory.
• The Jay Treaty of 1794 between the United States and Great Britain is the conventional starting point: it created mixed commissions to settle boundary questions, pre-war debts and maritime seizures — the first modern instance of states agreeing in advance to submit live disputes to a third-party body.
• The Alabama Claims arbitration of 1872 made the practice respectable. Britain had allowed Confederate commerce raiders to be built in its shipyards; a five-member tribunal sitting in Geneva found it in breach of its neutrality obligations and awarded the United States US$15.5 million, which Britain paid.
  • The significance is the payment, not the award. A great power accepted an adverse ruling from a body it did not control, and complied with it.
• The Hague Peace Conferences of 1899 and 1907, convened at the initiative of Tsar Nicholas II, produced the Convention for the Pacific Settlement of International Disputes and with it the Permanent Court of Arbitration, which began work in 1902.

The Permanent Court of Arbitration is neither permanent nor a court

• The PCA is not a court and is not the ICJ, and the two are routinely confused because they share a city and a building — the Peace Palace, opened in 1913 with Andrew Carnegie’s money.
• What the PCA maintains is a panel of arbitrators, four nominated by each contracting party, plus an International Bureau providing registry services to tribunals the parties themselves constitute. There is no permanent bench; cases are heard by arbitrators picked from the list.
• It remains busy — the Enrica Lexie arbitration and the Indus Waters Court of Arbitration are both PCA-administered — and it handles investor-state and mixed disputes the ICJ is barred from touching. Its national groups nominate candidates for election to the ICJ bench.

The Permanent Court of International Justice, 1922–1946

• Article 14 of the Covenant of the League of Nations charged the League Council with formulating plans for a Permanent Court of International Justice — the first standing international tribunal with general jurisdiction, competent to hear any dispute states brought rather than one class of dispute.
• An Advisory Committee of Jurists drafted the Statute in 1920; the PCIJ held its inaugural sitting in 1922 at the Peace Palace, with eleven judges and four deputy judges, later fifteen.
• Between 1922 and 1940 it dealt with twenty-nine contentious disputes and twenty-seven advisory opinions, and its jurisprudence — Lotus, Mavrommatis on diplomatic protection, Chorzów Factory on reparation — is still cited today.
• The German invasion of the Netherlands in 1940 ended its work; it was formally dissolved in 1946, its members resigning together that January.
• Its real achievement was a demonstration rather than a ruling: a standing international court could sit, produce a coherent body of case law, and be complied with.

The choice made at San Francisco

• The Washington Committee of Jurists in April 1945 chose a new court rather than continuation, for political reasons: the PCIJ was a League organ, and several founding UN members, the Soviet Union above all, had no wish to inherit a body of the interwar order.
• The break is formal and the continuity substantive. Article 92 of the Charter provides that the Court functions under a Statute “based upon the Statute of the Permanent Court of International Justice”. Article 36(5) carried existing PCIJ optional-clause declarations over, and Article 37 transferred PCIJ compromissory clauses.
• The Statute is annexed to the Charter and forms an integral part of it (Article 92), so every UN member is automatically a party to it and all 193 member states have standing to appear without any separate ratification.
• Article 93(2) allows a non-member to become a party on conditions set by the Assembly on the Council’s recommendation — Switzerland until it joined in 2002, Japan before 1956, and Liechtenstein, San Marino and Nauru before their own admission.

Where the Court sits in the United Nations system

• The ICJ is the principal judicial organ of the United Nations (Article 92) and the only principal organ not seated in New York — the Peace Palace at The Hague has been its seat since 1946, though Article 22 permits it to sit elsewhere.
• It has no hierarchical authority over the other organs. It is not a supreme court of the UN system; the Assembly and the Council are not its subordinates, and their resolutions are not appealable to it.
• Nor does it possess a general power of judicial review over the acts of the political organs. The point has been argued twice at length.
  • In the Namibia advisory opinion (1971) the Court stated plainly that it does not possess powers of judicial review or appeal over the decisions of the organs concerned — while nonetheless examining, in the course of answering, whether the resolutions in issue had been validly adopted.
  • In the Lockerbie cases (1992) Libya asked to be shielded from Security Council resolution 748. The Court declined provisional measures, holding that under Article 103 Charter obligations prevail; whether the Council’s determinations are reviewable at all was argued in separate opinions and never resolved.
• The Court can review incidentally, not directly. It may decide whether an organ acted within its powers in order to answer a question properly before it, as in Certain Expenses (1962), but no procedure exists for annulling a resolution.

The Court is not the International Criminal Court

Conflating the two is the commonest error on this subject, and the distinction is complete.

	International Court of Justice	International Criminal Court
Founding instrument	Statute annexed to the UN Charter, 1945	Rome Statute, in force 1 July 2002
Status	Principal organ of the United Nations	Independent treaty body, not a UN organ
Parties before it	Only states (Article 34)	Only natural persons
Subject matter	Any question of international law between states	Genocide, crimes against humanity, war crimes, aggression
Membership	All 193 UN members	About 125 states parties; not the US, China, India or Russia
Output	Judgments and advisory opinions	Convictions, acquittals, arrest warrants

• The two can look at the same events from opposite ends. Alleged genocide in Myanmar or Gaza generates an ICJ case about a state’s responsibility and, separately, ICC proceedings about an individual’s criminal liability.

The bench: composition, election and internal structure

The Statute’s provisions on who sits and how they get there are the Court’s answer to a hard design problem — how to build a tribunal that states will trust to judge them, out of judges whom states themselves choose.

Fifteen judges, nine years, staggered

• The Court consists of fifteen judges, described in Article 2 as “a body of independent judges”, and under Article 3 no two may be nationals of the same state.
• Judges serve nine-year terms and are re-eligible — Peter Tomka of Slovakia, on the bench since 2003, and Ronny Abraham of France, since 2005.
• Five seats fall vacant every three years, so the bench renews by thirds and never turns over at once — continuity of jurisprudence was the design intention, and the Court’s positions shift slowly.
• A judge elected to fill a casual vacancy serves only the remainder of the predecessor’s term. Mahmoud Hmoud of Jordan took the seat vacated when Nawaf Salam resigned in January 2025; Phoebe Okowa of Kenya, elected in November 2025, serves out Abdulqawi Yusuf’s term to February 2027.

How judges are elected — the double election with no veto

• Judges are elected by the General Assembly and the Security Council, voting separately and simultaneously in different chambers of the same building. A candidate needs an absolute majority in both — currently 97 votes in the Assembly and 8 in the Council.
• The permanent members have no veto in this election. Article 10(2) provides that any majority in the Council suffices “without any distinction between permanent and non-permanent members”. The Council is a second electorate here, not a gatekeeper.

Nomination by national groups, not by governments

• Candidates are not nominated by governments. Under Articles 4 to 6 nominations are made by the national groups in the Permanent Court of Arbitration — panels of up to four jurists each state maintains on the PCA list.
• A group may nominate at most four persons, no more than two of its own nationality, and Article 6 recommends prior consultation of the highest courts, law faculties and academies of law.
• The purpose is insulation — judges proposed by jurists rather than by foreign ministries, so that the bench is not a collection of national representatives.
• It works imperfectly. National groups are appointed by governments, usually mirror their preference, and elections turn on the same vote-trading and bloc arithmetic as any other UN election. The insulation is real at nomination and thin at election.

Qualifications and the representation requirement

• Under Article 2 a judge must be of high moral character and either possess the qualifications required in their own country for the highest judicial offices or be a jurisconsult of recognised competence in international law.
  • The second limb is why the bench mixes national judges with academics and government legal advisers — Hilary Charlesworth and Dire Tladi came from the academy, Bogdan Aurescu from government.
• Article 9 requires that in the body as a whole “the representation of the main forms of civilisation and of the principal legal systems of the world should be assured”.
  • This binds the bench collectively, not any seat, and is why the Court has always included judges from the common law, the civil law, Islamic law and formerly the socialist tradition. It is also the textual hook for the regional distribution, though Article 9 says nothing about regions.

The regional convention and the seat that was lost

• Seats are apportioned by an informal convention never written into any instrument, tracking the Assembly’s regional groups.

Regional group	Seats
African states	3
Asia-Pacific states	3
Latin American and Caribbean states	2
Eastern European states	2
Western European and other states	5

• Alongside it ran a second and stronger convention: that each permanent member would always have a judge on the bench. It held from 1946 without exception, and it was never a rule — nothing in the Statute or Charter mentions the permanent members.
• It broke in November 2017. India’s Dalveer Bhandari, seeking re-election, and the United Kingdom’s sitting judge Christopher Greenwood contested the last vacant seat. Through repeated rounds Bhandari held the General Assembly and Greenwood the Security Council, and neither could be elected without both.
  • With no veto available the Council could not impose its preference, and attrition favoured the candidate with the larger electorate behind him.
  • Before a further round the United Kingdom withdrew Greenwood’s candidacy, and Bhandari was elected with the Assembly’s absolute majority and the Council’s unanimous support.
  • The Court was left without a British judge for the first time since 1946, and the permanent members’ hold on the bench was shown to rest on votes, not entitlement.
• The convention eroded further when Russia lost its seat in the 2023 election — the first time since 1946 that no Russian or Soviet judge sat, and the first outright defeat of a permanent member’s candidate.

The bench is no longer the property of the great powers, and it was a routine election, not a reform, that made the point.

Judges ad hoc

• Article 31 allows a party with no judge of its nationality on the bench to choose a judge ad hoc for that case alone, so a bench may sit at sixteen, or seventeen where neither party has a national judge.
• The rationale is confidence: a state submits more readily to a bench containing someone certain to understand its legal system and argument. The device was inherited from arbitration, where each side appoints an arbitrator.
• The standing criticism is empirical. Judges ad hoc vote for the appointing state with near-perfect regularity — a licence for one guaranteed vote rather than a guarantee of understanding. The single dissent in Jadhav, carried 15 votes to 1, was cast by Judge ad hoc Tassaduq Jillani, appointed by Pakistan.

Independence, incompatibility and removal

• Article 16 bars a judge from any political or administrative function or other professional occupation; Article 17 bars sitting in any case in which the judge previously acted as agent, counsel or advocate.
• Article 18 provides the only route to removal: dismissal only if, in the unanimous opinion of the other members, the judge has ceased to fulfil the required conditions. It has never been used.
• Members enjoy diplomatic privileges and immunities (Article 19), and salaries fixed by the Assembly may not be decreased during a term (Article 32) — the guarantee against pressure through pay.

President, Vice-President and how the Court decides

• The Court elects a President and Vice-President from among its members for three-year terms, by secret ballot and absolute majority, both re-eligible. The President presides over all sittings, directs the Court’s work, and has a casting vote where the Court divides equally.
• The present holders are President Iwasawa Yuji of Japan, who assumed the office on 3 March 2025, and Vice-President Julia Sebutinde of Uganda, since 6 February 2024.
• The casting vote has changed an outcome once, with lasting damage. In the South West Africa cases (1966) the Court divided seven to seven on whether Ethiopia and Liberia had a legal interest entitling them to complain of South Africa’s administration of the mandate; the casting vote of President Sir Percy Spender decided that they did not, and the claims were rejected without any ruling on apartheid.
  • The verdict was read across Africa and Asia as a Western bench shielding a white-minority regime on a technicality. The Assembly terminated South Africa’s mandate by resolution 2145 of 27 October 1966, several judges were not returned at the next election, and the Court’s Global South caseload collapsed for a generation.
• The quorum is nine (Article 25), decisions are by majority of the judges present, and the President’s casting vote breaks a tie.
• Any judge may append an opinion — a dissenting opinion where the judge voted against the operative decision, a separate opinion where the judge agreed on the result but not the reasoning, and a declaration stating agreement or disagreement without full reasoning.
  • The practice is inherited from the common law and unusual among civil-law courts. The losing arguments are published alongside the winning ones, and today’s dissent is frequently tomorrow’s majority.

Chambers, the Registry and the languages

• The Court normally sits as a full court, but the Statute provides three smaller formations.
  • A Chamber of Summary Procedure of five judges, constituted annually under Article 29, has never given a judgment.
  • Ad hoc chambers under Article 26(2), formed for a particular case with the parties’ approval of composition — the first in the Gulf of Maine case (Canada/United States, 1984), followed by a handful of frontier disputes.
  • Standing chambers under Article 26(1) — the Court created a Chamber for Environmental Matters in 1993, reconstituted it annually until 2006, and then stopped, because no state ever asked to use it. States prefer the full Court’s authority to a specialist chamber’s expertise.
• The Registry, headed by a Registrar elected for seven years, is the Court’s permanent administrative organ — the ICJ is the only principal organ with its own independent administration.
• The official languages are French and English (Article 39). Every judgment is delivered in both, and the Court decides at the outset which text is authoritative.

Contentious jurisdiction: who may sue, and on what basis

The Court’s jurisdiction is twofold — contentious, deciding legal disputes between states with binding effect, and advisory, answering legal questions put by authorised UN bodies. The contentious side is constrained twice over: by who may appear at all, and by the requirement of consent.

Only states may be parties

• Article 34(1) of the Statute is categorical: “Only states may be parties in cases before the Court.” This is jurisdiction ratione personae, and it excludes absolutely:
  • individuals, however grave the wrong done to them, and corporations, including those whose assets a state has seized;
  • non-governmental organisations;
  • international organisations, the United Nations itself included;
  • peoples and self-determination movements, and entities whose statehood is contested.
• Under Article 34(2) the Court may request information from public international organisations, and they may furnish it unasked. They may inform the Court; they may not litigate before it.
  • The anomaly is stark. The UN has international legal personality and can bring an international claim — the Court said so in Reparation for Injuries (1949) — yet cannot bring that claim to its own principal judicial organ. Its only route is an advisory request.

Diplomatic protection: the route in for people and companies

• Diplomatic protection is the doctrine by which a state espouses the claim of its own national and thereby converts it into the state’s own. The injury to the national is treated as an injury to the state of nationality.
• The classic statement is the PCIJ’s in Mavrommatis Palestine Concessions (1924): in taking up its subject’s case a state is asserting its own right to ensure respect for international law in the person of its nationals.
• The consequences are severe for the individual: the state has complete discretion whether to espouse a claim, may settle or discontinue it, and any compensation goes to the state, which need not pass it on.
• Two conditions apply: nationality of the claim, and exhaustion of local remedies.
• Nottebohm (1955) supplied the genuine link requirement — Nottebohm, resident in Guatemala for thirty-four years, took Liechtenstein nationality in 1939, and the Court held that nationality conferred without a genuine connection of attachment is not opposable to third states.
• Barcelona Traction (Belgium v. Spain, 1970) settled the position of companies: the right to protect belongs to the state of incorporation — Canada — and Belgian shareholders had no independent claim.
• Diallo (Guinea v. DRC) modernised the doctrine, extending it in 2007 to internationally guaranteed human rights and producing a rare compensation award in 2012.
• Jadhav is diplomatic protection in operation. Kulbhushan Jadhav could not go to The Hague himself; India brought the claim as a breach of its own rights under the Vienna Convention on Consular Relations.

The Court hears people’s cases only when a government decides to make them its own, which is a design choice about sovereignty, not an oversight.

Jurisdiction rests on consent: the four bases

• There is no compulsory jurisdiction in the domestic sense. A domestic court’s power over a defendant does not depend on the defendant agreeing to be sued; the ICJ’s does. Consent is the foundation of the Court’s entire contentious competence.
• Consent may be given before or after a dispute arises, generally or specifically. Article 36, read with practice, yields four bases.

Basis	When consent is given	Form	Illustration
Special agreement (compromis)	After the dispute arises	Joint submission of an agreed question	Gulf of Maine; Gabčíkovo-Nagymaros
Compromissory clause	Before, in a treaty	Article 36(1) — matters “specially provided for”	Genocide Convention Art. IX; the Consular Relations Optional Protocol
Optional clause	Before, unilaterally	Article 36(2) declaration, usually reserved	Nicaragua; Aerial Incident (1999)
Forum prorogatum	After the application is filed	Respondent’s later acceptance	Corfu Channel; Certain Criminal Proceedings in France

Special agreement

• Two or more states conclude a special agreement, or compromis, referring an existing dispute to the Court and jointly defining the question. There is no applicant and no respondent — the parties appear as co-submitters.
• It is the most effective basis by a distance. Consent is explicit, contemporaneous and mutual, and states that have gone to that trouble have already decided to accept the outcome. Compliance here is close to complete, and the category is dominated by boundary and delimitation cases.

Compromissory clause in a treaty

• Article 36(1) extends jurisdiction to “all matters specially provided for in the Charter of the United Nations or in treaties and conventions in force”. A compromissory clause is the provision by which parties agree in advance that disputes about a treaty go to the Court.
• Several hundred treaties contain one; two do most of the current work.
  • Article IX of the Genocide Convention (1948) — the basis of Bosnia v. Serbia, Croatia v. Serbia, The Gambia v. Myanmar, Ukraine v. Russia and South Africa v. Israel. States that reserved against it, the United States among them, cannot be sued under the Convention.
  • The Optional Protocol to the Vienna Convention on Consular Relations (1963) — the basis of LaGrand, Avena and Jadhav. The United States withdrew from it in 2005, after losing Avena.
• The trend runs against the Court. These cases produce weaker compliance than special agreements, because one party is sued on a consent given years earlier for another purpose; and since the 1970s treaty drafters have increasingly provided for arbitration instead.

The optional clause: “compulsory”, “voluntary” and “optional” jurisdiction

Three names attach to overlapping ideas here, and clearing the confusion comes before anything else.

• Article 36(2) allows a state to declare that it recognises, in relation to any other state accepting the same obligation, the Court’s jurisdiction as compulsory ipso facto, in all legal disputes concerning treaty interpretation, questions of international law, the existence of a breach, and reparation.
• “Compulsory jurisdiction” is the Statute’s own phrase and describes the effect: once two states have declared, either may sue the other without fresh agreement, and the respondent cannot refuse to appear.
• “Optional clause” describes the act, which is why Article 36(2) has carried that name since the PCIJ; “voluntary jurisdiction” is the same mechanism seen from the state’s side.
• All three phrases point at one thing, and the paradox dissolves once it is seen: acceptance is voluntary, and what is voluntarily accepted is compulsion.
• A declaration is deposited unilaterally with the Secretary-General and creates a network of consent: any two declaring states are subject to the Court’s jurisdiction as between themselves, so far as their declarations overlap.

Take-up is thin and getting no better.

• About seventy-five states out of 193 have a declaration in force — under forty per cent, and a smaller share than under the PCIJ, where roughly two-thirds of a far smaller League membership had accepted.
• Only one permanent member has a declaration: the United Kingdom, repeatedly narrowed, most recently to exclude nuclear-weapons disputes after the Marshall Islands litigation.
  • France withdrew in 1974 after the Nuclear Tests cases over atmospheric testing in the Pacific; the United States withdrew in 1985, effective 1986, during Nicaragua; China and Russia have never declared.
• Every state that could plausibly be a serious respondent has therefore either never accepted the clause or reserved its way out of the categories that matter.

Reciprocity narrows every declaration to the narrower of the pair.

• Article 36(2) operates “in relation to any other state accepting the same obligation“, so a respondent may invoke the applicant’s reservations against the applicant, and jurisdiction exists only where two declarations overlap.
• The narrowest declaration in any pair governs. A state with broad acceptance is protected by its opponent’s exclusions, and every new reservation anywhere degrades the system for everyone.
• Reciprocity runs to timing too. In Right of Passage the Court accepted that Portugal could sue India days after depositing its own declaration; states responded with twelve-month waiting-period reservations — India’s among them — to defeat exactly that manoeuvre.

Self-judging reservations, and how they rebounded.

• The Connally reservation, attached by the United States in 1946, excluded “disputes with regard to matters which are essentially within the domestic jurisdiction of the United States of America as determined by the United States of America”.
• The final clause makes the reservation self-judging — the reserving state decides for itself whether the exclusion applies, which appears to leave the Court no role in policing its own jurisdiction, against the principle of compétence de la compétence in Article 36(6).
• It boomeranged twice, exactly as reciprocity predicts.
  • In Certain Norwegian Loans (France v. Norway, 1957) France had a self-judging reservation of its own; Norway invoked France’s reservation against France and the Court held it had no jurisdiction. The applicant was defeated by its own words.
  • In Interhandel (Switzerland v. United States, 1959) the United States pleaded the Connally reservation; the Court avoided ruling on its validity and disposed of the case on failure to exhaust local remedies.
• Judge Sir Hersch Lauterpacht argued in both that a self-judging reservation is invalid, and that being an essential condition of the declaration its invalidity destroys the whole declaration. The Court has never decided the question.
• A deeper problem is structural: a declaration may be withdrawn or modified, so a state can accept compulsory jurisdiction while it is convenient and leave when it is not. Nicaragua is the demonstration — the United States tried to modify its declaration three days before Nicaragua filed, the Court held the attempt ineffective, and Washington withdrew.

Forum prorogatum

• Forum prorogatum is jurisdiction founded on the respondent’s consent given after the application is filed, expressly or by conduct.
• It was established in Corfu Channel (United Kingdom v. Albania, 1948): Albania had neither declaration nor treaty clause, but wrote accepting jurisdiction for that case, and the Court held that consent needs no particular form.
• Since 1978 the Rules of Court, Article 38(5), have regularised it — an application against a non-consenting state is not entered in the General List until that state consents, which ended the abuse of publicity-seeking filings.
• Consent is rare but real: France consented in Certain Criminal Proceedings in France (2003), the first under the revised rule, and Djibouti v. France (2008) proceeded the same way.

Provisional measures under Article 41

• Under Article 41 the Court may indicate provisional measures to preserve the rights of either party pending judgment. They are its emergency jurisdiction and the source of most of its contemporary visibility.
• The classic doubt was whether they bind. Article 41 says the Court may “indicate” measures which “ought to be taken”; for fifty years states treated them as recommendations.
• LaGrand (Germany v. United States, 27 June 2001) settled it. The Court had ordered the United States to prevent Walter LaGrand’s execution pending judgment; he was executed the same day. The Court then held for the first time that orders indicating provisional measures are binding, since a power to preserve rights is meaningless if it can be disregarded.
• The threshold has three settled elements.
  • Prima facie jurisdiction — the basis invoked need only appear to afford one.
  • Plausibility — the rights asserted must be plausible, with a link to the measures sought; formalised in Belgium v. Senegal (2009) and now the pivot of every order.
  • Urgency and irreparable prejudice — a real and imminent risk of harm no later judgment could repair.
• They now carry the Court’s public profile, because they issue in weeks while merits take a decade. Ukraine v. Russia, order of 16 March 2022, directed Russia to suspend its military operations and was ignored; The Gambia v. Myanmar, 23 January 2020, directed Myanmar to prevent genocidal acts against the Rohingya; South Africa v. Israel produced four orders in 2024.
• Compliance with provisional measures is far worse than with judgments — studies place it around half and falling, the predictable result of their being sought precisely where the respondent least intends to cooperate.

Intervention by third states

• Article 62 lets a state with an interest of a legal nature which may be affected seek permission to intervene; requests are usually refused, Nicaragua in El Salvador/Honduras (1990) being the leading success.
• Article 63 gives a state party to a multilateral convention whose construction is in question a right to intervene on a simple declaration, and the construction adopted then binds it equally.
• Article 63 intervention was rare for seventy years and has become a mass phenomenon.
  • In Ukraine v. Russia, thirty-three states filed declarations and the Court admitted thirty-two at the preliminary-objections stage — by a wide margin the largest intervention in the history of international adjudication.
  • The Gambia v. Myanmar proceeds with eleven states intervening, beginning with the Maldives and a joint declaration by Canada, Denmark, France, Germany, the Netherlands and the UK.
  • South Africa v. Israel has drawn declarations from Türkiye, Spain, Mexico, Chile, Colombia, Ireland, Cuba, Palestine, Brazil and Belgium.
• It matters twice over: a bilateral case becomes a collective proceeding about a treaty’s meaning, and states now treat litigation as a foreign-policy instrument. The counter-argument is that this politicises the courtroom and lets the respondent claim it faces a coalition, not a case.

Preliminary objections, non-appearance and default

• A respondent contesting jurisdiction or admissibility raises preliminary objections, decided in a separate judgment that suspends the merits — typically attacking the jurisdictional basis, the existence of a dispute, standing, or exhaustion of local remedies.
  • Preliminary objections are where most cases die. The Aerial Incident case (2000) and the Marshall Islands cases (2016) both ended there with no examination of the substance.
  • The Marshall Islands cases were dismissed because no “dispute” existed at the date of application — the respondents had not been shown to be positively opposed to the claim. Against the United Kingdom the point was carried by the President’s casting vote, and the reasoning is widely criticised as an unduly formal way of avoiding a politically impossible subject.
• Non-appearance.Article 53 provides that where a party does not appear the other may ask the Court to decide in its favour, but the Court must first satisfy itself that it has jurisdiction and that the claim is well founded in fact and law. Non-appearance neither stops the case nor produces an automatic win.
  • Iran did not appear in Tehran Hostages (1980); the United States withdrew from Nicaragua after the 1984 jurisdiction judgment; Russia has not participated on the merits in the cases now against it. The Court proceeded in each.

Advisory jurisdiction

• Article 96(1) allows the General Assembly and Security Council to request an advisory opinion on any legal question; Article 96(2) allows other organs and specialised agencies, when authorised, to ask about questions arising within their activities.
• Roughly thirty opinions have been given by the ICJ and twenty-seven by the PCIJ; only states and authorised organisations may take part.
• The limit in Article 96(2) has teeth. In 1996 the Court answered the Assembly’s question on the legality of the threat or use of nuclear weapons but refused the parallel request from the World Health Organization as ultra vires — nuclear legality is not a question arising within the WHO’s activities. An agency’s competence limits the question it may ask.
• Opinions are not binding on anyone. Their authority is persuasive and considerable — statements of the law by the principal judicial organ, reasoned and usually near-unanimous.
• The discretion to decline is real and almost never exercised. Only “compelling reasons” justify refusal, and refusal has happened once in the history of both courts — the PCIJ in Eastern Carelia (1923), because answering would have decided a dispute involving a non-consenting non-member.
• Advisory opinions have become a strategic instrument. They need no respondent’s consent, cost a coalition nothing but Assembly votes, and produce an authoritative statement that reshapes the political argument. Chagos, Palestine and climate change are all coalitions unable to sue going to the Assembly instead.

Opinion	Year	What it established
Reparation for Injuries	1949	The UN has international legal personality; functional protection of its agents
Certain Expenses	1962	Peacekeeping costs are “expenses of the Organization”; implied powers of the organs
Namibia	1971	South Africa’s presence unlawful; abstention is not a veto; no power of judicial review
Western Sahara	1975	No ties of sovereignty displacing self-determination
Nuclear Weapons	1996	Generally contrary to humanitarian law; no conclusion in an extreme case of self-defence; duty to pursue disarmament in good faith
Wall	2004	Contrary to international law; erga omnes consequences and third-state duties of non-recognition
Kosovo	2010	The declaration of independence did not violate general international law
Chagos	2019	Decolonisation of Mauritius not lawfully completed; a continuing wrongful act
Occupied Palestinian Territory	2024	Continued presence unlawful; obligation to end it as rapidly as possible
Climate Change	2025	Obligations under treaty and custom; erga omnes; breach engages full reparation
Israel and the UN presence	2025	Duty to facilitate relief and cooperate with UN bodies; UNRWA not presently replaceable

The climate advisory opinion of 23 July 2025

• The request came by General Assembly resolution 77/276 of 29 March 2023, adopted by consensus with 132 co-sponsors, initiated by Vanuatu and driven from the outset by Pacific law students — an origin without parallel in the Court’s history.
• Two questions were put: what obligations states have to protect the climate system for present and future generations, and what legal consequences follow where a state has caused significant harm.
• Participation was the largest ever — 91 written statements and 107 oral statements at hearings in December 2024, with 96 states and 11 organisations taking part and several small island states appearing for the first time.
• The opinion was unanimous, which for fifteen judges from five continents on the most contested question in world politics is itself the headline.
• The holdings.
  • Obligations arise not only under the climate treaties — the UNFCCC, Kyoto and Paris — but independently under customary international law, including the duty to prevent significant transboundary harm and obligations under human rights law. States outside the treaty regime are not outside the law.
  • 1.5°C is the primary temperature goal against which conduct is assessed.
  • The production, consumption, licensing and subsidising of fossil fuels may constitute an internationally wrongful act attributable to the state.
  • The obligations are erga omnes — owed to the international community as a whole, because a stable climate is a common good — so any state may invoke another’s responsibility for their breach without proving specific injury.
  • Breach engages the full law of state responsibility: cessation, guarantees of non-repetition and full reparation, including compensation, where a sufficiently direct and certain causal link is shown. Attribution of harm to particular states is scientifically and legally possible case by case.
• Why it matters. For small island states it converts a moral claim into a legal one; for domestic and regional courts it supplies an authoritative statement of the law, and litigants began citing it within weeks. The climate regime itself is a separate subject; what belongs here is what the Court did to it.
• The caveat. It is advisory and binds no one; it creates no forum in which a large emitter can actually be sued, since none of the biggest accepts compulsory jurisdiction; and causation remains the hard problem — the Court said a link can be established, not that establishing one will be easy.

The climate opinion changed what states can be argued to owe; it did not change who can be made to pay.

Judgments, enforcement and the compliance record

• Article 59 limits a judgment’s binding force to the parties and that case — formally there is no precedent, though the Court cites itself constantly.
• Article 60: judgments are final and without appeal. Two narrow remedies exist — interpretation where the parties dispute what was decided, and revision under Article 61 on a decisive new fact, never after ten years. No application for revision has ever succeeded.
• Article 94(1): every member undertakes to comply with the decision in any case to which it is a party. Article 94(2): on non-performance the other party may have recourse to the Security Council, which may recommend or decide upon measures to give effect to the judgment.

Why Article 94(2) is an empty remedy

• The veto sits inside the enforcement clause. A permanent member that loses can block enforcement against itself, and against a client. The veto’s own law and record belong with the Council itself; what matters here is its effect on judgments.
• Nicaragua is the demonstration. The Court held in 1986 that United States support for the contras and the mining of Nicaraguan harbours breached the customary prohibitions on force and intervention. Nicaragua went to the Council under Article 94(2); the United States vetoed the enforcement resolution and then withdrew its optional-clause declaration.
• The Council has never once adopted a measure under Article 94(2) in eighty years. The structural point is uncomfortable: the states most likely to be defendants in the cases that matter most are exactly those the enforcement machinery cannot reach.

The record is better than the theory predicts

• Most judgments are complied with. Studies of 1946 to 1987 put full compliance near eighty per cent; for 1987 to 2004 the figure is nearer sixty per cent, the later sample containing far more cases brought against unwilling respondents.
• Compliance is highest where consent was clearest. Judgments founded on a special agreement are complied with almost invariably, and boundary and maritime-delimitation judgments are implemented as a matter of course, often with joint demarcation commissions afterwards.
  • A state that agreed to be judged has already decided to accept the outcome. Litigation is a device for making a concession survivable at home — the government did not give ground, the Court took it.
  • Nigeria’s implementation of the Bakassi judgment (2002), transferring the peninsula to Cameroon, is the standing example of a hard territorial loss actually executed.
• The failures are concentrated and conspicuous — Corfu Channel (Albania paid only in 1996, forty-seven years late), Tehran Hostages, Nicaragua, Avena, and the provisional-measures orders outstanding against Russia, Myanmar and Israel.
• Both propositions are true at once: the Court works reliably on the disputes states are willing to settle and unreliably on the disputes that shape world politics. A high compliance rate across all cases is partly an artefact of which cases reach judgment at all.

Case	Year	What it settled
Corfu Channel	1949	The first case; state responsibility for knowingly allowing territory to be used to harm others
Reparation for Injuries	1949	Legal personality of international organisations
Nottebohm	1955	Genuine link requirement for nationality
Right of Passage	1960	Local custom between two states can create rights
South West Africa	1966	Claims rejected on standing by the casting vote; the Court’s credibility crisis
North Sea Continental Shelf	1969	How customary law is formed — practice plus opinio juris
Barcelona Traction	1970	Obligations erga omnes; shareholders cannot claim for the company
Nuclear Tests	1974	Unilateral declarations can create binding obligations
Tehran Hostages	1980	Inviolability of diplomatic premises and personnel; attribution of endorsed conduct
Nicaragua	1986	Custom survives alongside treaty law; non-intervention; the “effective control” test
Gabčíkovo-Nagymaros	1997	Sustainable development; necessity; treaties survive changed circumstances
LaGrand	2001	Provisional measures are binding; Article 36 VCCR creates individual rights
Avena	2004	Review and reconsideration as the remedy for consular violations
Bosnia v. Serbia	2007	The Convention imposes an obligation to prevent; attribution failed on effective control
Pulp Mills	2010	Environmental impact assessment is customary international law
Whaling in the Antarctic	2014	Japan’s programme not “for purposes of scientific research”
Jadhav	2019	Breach of Article 36 VCCR; no espionage exception

The caseload today

• 202 cases had been entered in the General List between 22 May 1947 and mid-2026, contentious and advisory together, roughly a third of them in the last two decades.
• Around two dozen cases are pending, the largest standing docket in the Court’s history, and its composition has changed as much as its size — genocide, environmental and human-rights claims now sit alongside the traditional boundary work.
• The Court is busier than ever at the moment multilateral politics is most obstructed — either law filling the space politics has vacated, or states litigating because nothing else is left.

India and the World Court

India’s relationship with the Court is unusually instructive. It has been applicant and respondent, has won and lost on jurisdiction, and its own declaration is a case study in how states use reservations to keep the Court at a safe distance.

The cases

• Right of Passage over Indian Territory (Portugal v. India, 12 April 1960). Portugal claimed passage across Indian territory to its enclaves of Dadra and Nagar Haveli, cut off after their absorption in 1954.
  • The Court held that a local custom binding only two states can exist — practice accepted as law over more than a century created a right of passage for private persons, civil officials and goods, but not for armed forces, armed police or arms and ammunition.
  • India then won on the substance: no breach, its 1954 refusal being justified by the disturbances and within its power to regulate passage.
• The Trial of Pakistani Prisoners of War case (1973), over the proposed trial in Bangladesh of 195 prisoners held after the 1971 war, was removed from the List in December 1973 at Pakistan’s request after a negotiated settlement.
• Aerial Incident of 10 August 1999 (Pakistan v. India, 21 June 2000). Pakistan sued over the shooting down of a naval aircraft, invoking the 1928 General Act for the Pacific Settlement of International Disputes, the parties’ declarations, and Article 36(1).
  • The Court held by fourteen votes to two that it had no jurisdiction: India was not bound by the 1928 General Act, the Charter created no independent basis, and India’s declaration excluded disputes with Commonwealth states and disputes under multilateral treaties unless all parties were before the Court.
  • It confirmed that attaching reservations is a recognised practice which it was bound to apply as written, while reminding both governments of their duty to settle disputes peacefully under the Simla Agreement (1972) and the Lahore Declaration (1999).
  • This is the case that shows what India’s reservations are for — to keep the India–Pakistan file out of The Hague, and they worked as designed.
• The Marshall Islands case (2016), one of three brought against India, the United Kingdom and Pakistan over the duty to negotiate nuclear disarmament, was dismissed for absence of a dispute.
• Jadhav (India v. Pakistan), 17 July 2019, by fifteen votes to one. Kulbhushan Jadhav was arrested in Pakistan in March 2016 and sentenced to death by a military court in April 2017; Pakistan neither informed him of his consular rights nor granted India access.
  • Jurisdiction did not rest on the optional clause at all. India invoked Article I of the Optional Protocol to the Vienna Convention on Consular Relations, so India’s Commonwealth reservation was simply irrelevant.
  • The Court rejected Pakistan’s three defences: Article 36 contains no espionage exception; customary law does not displace an express treaty provision; and the India–Pakistan agreement on consular access of 2008 does not derogate from the Convention.
  • It found breaches of Article 36(1) on three counts — failure to inform Jadhav of his rights, three weeks’ delay in notifying India in violation of “without delay”, and denial of consular access.
  • Remedies. The Court refused India’s requests for annulment and release, and ordered effective review and reconsideration of the conviction and sentence by means of Pakistan’s choosing, with a continued stay of execution as an indispensable condition.
  • The sole dissent was Judge ad hoc Tassaduq Jillani, appointed by Pakistan. Pakistan responded with the ICJ (Review and Reconsideration) Ordinance of 2020, enacted as an Act in 2021; Jadhav remains in custody and the review has not been effectively completed.

India won in 2019 because a treaty’s own compromissory clause bypassed the restrictive declaration India had written to protect itself.

India’s acceptance of compulsory jurisdiction

• India first accepted the optional clause in 1940 as British India, and after independence declared in 1956, 1959 and, most consequentially, on 18 September 1974 — the declaration revoked and replaced by a new one of 27 September 2019, which keeps the same architecture and extends it.
• The excluded categories, taken together, are among the most extensive of any declaring state.

Reservation	What it defends against
Parties have agreed another method of settlement	Keeps bilateral machinery — Simla, the Indus Waters Treaty — out of the Court
Disputes with Commonwealth states, present or former	The Pakistan exclusion; defeated Pakistan in 2000
Matters essentially within India’s domestic jurisdiction	Internal security, citizenship, Kashmir’s constitutional status
Hostilities, self-defence, and action under Security Council resolutions	Every use of force by Indian forces
States accepting jurisdiction less than twelve months earlier	The Right of Passage manoeuvre — declaring in order to sue
Disputes under League of Nations treaties absent special agreement	Closes the route Pakistan tried through the 1928 General Act
Multilateral treaties unless all parties are before the Court	Blocks selective litigation under widely ratified conventions
States with no diplomatic relations with, or not recognised by, India	Removes a channel for unrecognised entities
Disputes with non-sovereign states or territories	The same, at the level of statehood
Territory, boundaries, maritime zones, airspace, delimitation	The core exclusion — Kashmir, Aksai Chin, Arunachal Pradesh, every frontier
Disputes arising from facts prior to the declaration	Everything historic, Partition included

• The honest assessment. Once territory, frontiers, hostilities, Commonwealth states, multilateral treaties, prior facts and domestic matters are removed, very little that could realistically be litigated against India remains inside the acceptance.
• Two defences are serious. Every state with something to lose does the same — the United Kingdom’s declaration is heavily reserved, and China, Russia, the United States and France do not accept the clause at all. And a state with live territorial disputes with two nuclear-armed neighbours cannot rationally expose them.
• The tension is equally real. India argues for a rules-based order and reformed multilateralism, and insulating itself from the Court on everything contentious carries an argumentative cost. India has never brought a case founded on its own declaration — Jadhav came through a treaty protocol.
• India is not hostile to third-party settlement as such: it has appeared before a PCA Court of Arbitration and a Neutral Expert on the Indus Waters Treaty, before the Annex VII tribunal in the Bay of Bengal delimitation with Bangladesh (2014), whose adverse elements it implemented, and in the Enrica Lexie arbitration. India prefers arbitration it helps constitute to a court it does not control.

Indians on the bench

• Sir Benegal Narsing Rau, a principal draftsman of the Indian Constitution, served from 1952 until his death in 1953.
• Nagendra Singh served from 1973 to 1988 and was President from 1985 to 1988 — the only Indian to hold the office, presiding over the Nicaragua merits judgment.
• R.S. Pathak, a former Chief Justice of India, served from 1989 to 1991; Dalveer Bhandari, elected in April 2012 and re-elected in November 2017, sits until 2027.
• Neeru Chadha became the first Indian woman elected to the International Tribunal for the Law of the Sea in June 2017.

The case against the Court

• The consent problem is foundational. A tribunal whose jurisdiction depends on the defendant’s agreement is not a court in the domestic sense; it is arbitration with a permanent bench. The optional clause was meant to cure this and has not.
• Enforcement runs through a Council with a veto. Article 94(2) has never been used, and the states least reachable by it are those whose conduct most needs reaching. Compliance rests on reputation, reciprocity and self-interest.
• The Court is slow and expensive. Cases routinely take a decade; in South Africa v. Israel the reply is due in November 2027 and the rejoinder in May 2029, so a merits judgment before the 2030s is improbable. States wanting a question resolved in reasonable time go to arbitration.
• Selectivity and politicisation are alleged from both directions.
  • That it avoids the great powers, disposing of politically impossible cases on jurisdiction or admissibility — Marshall Islands, and the Nuclear Weapons opinion’s inability to conclude.
  • That it has become a forum for political campaigns dressed as legal claims, with mass interventions and compromissory clauses stretched to cover disputes they were never written for.
  • Two episodes cost it dearly: South West Africa (1966) cost a generation of Global South confidence, Nicaragua (1986) cost the United States, which has never returned to the optional clause.
• Fragmentation and the proliferation of tribunals. The ICJ is one court among many — ITLOS, the ICC, WTO dispute settlement, the European and Inter-American human rights courts, investor-state arbitration, ad hoc PCA tribunals.
  • The risk is inconsistent jurisprudence. The textbook clash is attribution: the ICJ requires “effective control” over specific operations before a state answers for an armed group (Nicaragua, reaffirmed in Bosnia v. Serbia), while the ICTY in Tadić (1999) adopted a looser “overall control” test. Neither can overrule the other.
  • The counter-argument is that proliferation is a symptom of success — specialised tribunals exist because there is now enough international law to need them, and the ICJ’s authority means its positions usually prevail anyway.
  • The Court now competes for cases with bodies that are faster, more specialised, and open to parties it cannot hear.

International law and international cooperation

The Court’s largest contribution is not any judgment but its role in the system judgments belong to. Whether that system is law at all, and whether it does any work, is the oldest argument in the discipline.

Where international law comes from

• Article 38(1) of the Statute is the canonical statement of the sources the Court applies, and by extension of the sources of international law itself.
  • International conventions, general or particular, establishing rules expressly recognised by the contesting states.
  • International custom, as evidence of a general practice accepted as law — the two elements being consistent state practice and opinio juris, the belief that the practice is legally required. North Sea Continental Shelf (1969) is the Court’s fullest analysis of how the two combine.
  • General principles of law recognised by civilised nations — good faith, estoppel, res judicata, reparation for wrongs.
  • Judicial decisions and the teachings of the most highly qualified publicists, expressly as a subsidiary means for determining rules, not as sources in themselves.
• Because custom has no legislature, someone has to say authoritatively what it is, and that is the Court’s structural function. Nicaragua on force, Barcelona Traction on erga omnes obligations, Pulp Mills on impact assessment and the 2025 climate opinion all convert diffuse practice into a statable rule.

Why states obey

• Reciprocity. Most international law is a network of mutual conveniences — diplomatic immunity, postal union, air navigation, treaty performance. A state that breaks the rule loses the benefit of others keeping it, and the loss usually exceeds the gain.
• Reputation. A state known to break treaties finds its future promises discounted, which raises the price of everything it wants to negotiate.
• Domestic incorporation. Much international law is enforced not internationally but by national courts, legislatures and administrations — which is where most compliance actually happens.
• Transaction costs. Institutionalist theory, most fully in Robert Keohane, argues that regimes persist because they reduce the cost of making and monitoring agreements and stabilise expectations. States comply because the institution is cheaper than the alternative, not because they are coerced.
• Legitimacy. Thomas Franck argued that rules exert a “compliance pull” proportionate to their perceived legitimacy — determinacy, symbolic validation, coherence, adherence to a rule hierarchy — so states obey rules they regard as rightly made.
• Internalisation. Harold Koh’s account of transnational legal process describes compliance as the end point of repeated interaction, interpretation and internalisation: norms argued over in international fora are absorbed into domestic legal and bureaucratic practice until obedience becomes automatic.

Austin’s objection, and the answers

• John Austin held that law properly so called is the command of a sovereign backed by sanction. International law has no sovereign, no legislature and no enforcement machinery, and is therefore not law but “positive international morality” — closer to rules of honour than to law.
• Three answers are standard.
  • H.L.A. Hart replied that Austin’s model fails even for domestic law, which is not a set of commands but the union of primary and secondary rules. International law lacks a developed rule of recognition — but it functions as law in every other respect, and its rules are treated as obligatory rather than merely prudent.
  • Sanction is not what makes most law work. Constitutional law binds the officials who control the sanctioning machinery; conventions operate without sanction at all. Coercion is the exceptional case of legal obligation, not its definition.
  • Enforcement here is horizontal, not vertical — countermeasures, retorsion, non-recognition, sanctions, exclusion from institutions, reputational cost. Decentralised and unevenly available, which makes it weak; not absent.
• Hans Morgenthau’s realist objection is harder. He did not deny that international law exists; he argued it is decentralised law of a primitive kind, obeyed where interests coincide, and shaped by the distribution of power on questions engaging vital interests. Law records settlements the powerful have accepted; it does not produce them.
• The honest middle position is neither the formalist’s nor the realist’s.
  • International law is real law — states argue in its vocabulary even while breaking it, and the effort spent on justification is itself evidence of obligation. No state claims a right of aggression; it claims self-defence.
  • It works in the routine, governing the enormous volume of ordinary interaction — trade, transport, communications, diplomacy, the sea — where compliance is near-total and invisible.
  • It constrains the weak more reliably than the strong, which is a real injustice and not an argument that it is not law.
  • It is the medium through which cooperation is conducted. Cooperation between sovereigns needs commitments that outlast the moment of making, and law is the only technology that supplies them. The alternative to imperfect international law is not perfect politics but unmediated power.

International law does less than its defenders claim and far more than its critics concede, and durable cooperation has no other language.

Assessment

• What the Court is good at. Boundary and maritime delimitation, treaty interpretation, state responsibility, the authoritative statement of custom, and advisory opinions that change the terms of a political argument. In these areas its judgments are complied with and its reasoning is the working material of international law.
• What it cannot do. Compel a great power, prevent a war, act on its own initiative, hear an individual, or enforce anything. Its one enforcement clause is vetoable.
• What has changed. It is busier and hears a wider range of subject matter than ever; mass intervention and coalition-driven advisory requests have made it an instrument of collective legal action. It has lost its monopoly to specialised tribunals, and its bench is no longer the great powers’ to assume.
• The reform proposals that would matter, in rough order of feasibility.
  • Wider acceptance of the optional clause and withdrawal of self-judging reservations — the change that would do most, and the least likely.
  • Access for international organisations to contentious proceedings, at minimum for the United Nations, which cannot presently sue in its own court.
  • Wider use of Article 96(2) authorisations, and a power for the Secretary-General to request opinions — proposed repeatedly since the 1990s and never adopted.
  • Procedural acceleration — shorter time limits, greater use of chambers, firmer control of pleadings — to make the Court competitive with arbitration on speed.
  • Enforcement outside the Council: coordinated countermeasures, Assembly action, and conditioning institutional benefits on compliance, none of which needs a Charter amendment.
• The realistic verdict. The Court is neither the embryonic world court its founders imagined nor the ornament its critics describe. It settles the disputes states are willing to have settled, states the law for everyone else, and is powerless exactly where power is the issue. That is smaller than the Charter promised and larger than its absence would leave.