The Disputes Between Armenia and Azerbaijan: The CERD Compromissory Clause as a One-way Ticket to Hague

Written by


In September 2021, Armenia and Azerbaijan lodged each against the other an application before the International Court of Justice (Court) (ArA and AzA). Both applications refer to alleged breaches of the 1966 UN Convention on the Elimination of All Forms of Racial Discrimination (CERD), using its compromissory clause (Article 22) to establish the Court’s jurisdiction, and request provisional measures (PM). Hearings on PMs took place on 14-15 and 18-19 October 2021.

Besides a short description, I do not assess the applications (and PM requests) as such; Mike Becker did a terrific job at this. Instead, through these applications I discuss the trend of using compromissory clauses to litigate a fragment of broader conflicts, engaging other international legal sources. In a recent study (Rivista di Diritto Internazionale, 2021) I made, essentially, three claims. By pushing broader conflicts through specific treaty-based allegations:

  1. States seek less to solve the conflicts, than use the Court to score political points (domestically) and broadcast remonstrations (internationally);
  2. this strategy is lawful, but erodes the Court’s reputation/credibility as dispute settler; and
  3. might contribute to the fall into disgrace of compromissory clauses in new treaties.

The background

Armenia and Azerbaijan are embroiled in a conflict that – in its current permutation – dates back to the Soviet Union’s finale. In 1988, the Armenian majority in the ethnic enclave of Nagorno-Karabakh (southwest Azerbaijan) asked the Soviet Union to transfer the territory to Armenia. The move sparked a violent struggle between the two Republics, which deepened after the Soviet Union dissolved in a flurry of secessions. In December 1991, a local referendum (boycotted by the Azerbaijani community) prompted a declaration of independence from Azerbaijan. Since then, and in spite of a 1994 ceasefire, the two countries have fought over the area, in waves of varying intensity.

In September 2020, war erupted between the two countries over the control of Nagorno-Karabakh, ending formally on 10 November 2020 with a Russia-brokered ceasefire. Julia Emtseva described the practical insignificance of UNSC Resolution 2353 of July 2020 (on a global ceasefire during the pandemic). Bernhard Knoll-Tudor and Daniel Mueller assessed the legalities of the hostilities and the ceasefire’s terms. Azerbaijan obtained seven Armenian-occupied districts, and Russian peacekeeping forces were deployed.

The conflict continuously reconfigured the region’s power set-up and ethnical make-up. Either State can claim that – at some point – the inhabitants of its ethnicity were oppressed and/or displaced by the other’s authorities. Reconciliation is a long-term project.

The applications

The CERD prohibits racial discrimination, particularly racial segregation and apartheid. States must also remove, prevent and fight racial discrimination within their jurisdictions. Armenia joined the CERD in 1993, Azerbaijan in 1996.

Armenia’s application accused Azerbaijan of CERD breaches against individuals of Armenian ethnic or national origins. Azerbaijan’s application is not a defense, nor a counterclaim, but a fresh specular application. Azerbaijan argues that Armenia, through ethnic-cleansing, displaced almost all Azerbaijanis from the territories it controls. Both applications refer to the situations in Nagorno-Karabakh and in the responding State’s territory (for instance, the establishment of a “Military Trophy Park” in Baku, which in Armenia’s view provides a degrading image of its soldiers).

The Armenia’s application describes at length (ArA ¶¶23-34) Azerbaijani crimes against Armenians predating the respondent’s accession to CERD. Azerbaijan also discusses matters predating CERD’s application to Armenia, for instance Armenia’s policy and practice of ethnic cleansing “between 1987 and 1994” (AzA ¶10).

Armenia accuses Azerbaijan of CERD breaches and wrongdoings under other sources, which are presented as racially motivated. For example, after listing public declarations that would reveal the Azerbaijani policy of ethnic cleansing, Armenia mentions conduct that is not CERD-specific:

  1. Consistent with this policy of ethnic cleansing and rhetoric of hate, … Azerbaijan has historically committed countless violations of international law, the majority of which were plainly racially motivated.

  2. During and in the wake of the September-November 2020 conflict alone, numerous graphic videos widely circulated on the internet show Azerbaijani forces murdering, torturing, and subjecting Armenian civilians and prisoners of war to cruel and inhumane treatment.

Armenia attracts under the rubric of “racial discrimination” allegations of war crimes, violations of human rights and of the ceasefire. The request for PMs emphasises “murder, torture and other abuse of Armenian prisoners of war, hostages and other detained persons”. Likewise, Azerbaijan grafts CERD allegations onto conduct that is distinctly wrongful under other sources. The request for PMs demands primarily that Armenia facilitate demining in the region.

It is fair to say that these applications use a CERD-shaped cookie cutter on an enormous sheet of cookie dough, spanning over wrongdoings that transcend racial discrimination. “We all know it,” the parties know it, the Court knows it. Is it a problem?

The Court’s jurisdiction

Each State challenges the Court’s jurisdiction over the other’s application. Invoking the negotiation requirement (Article 22 CERD, central to the Georgia v. Russia outcome) requires creative lawyering. Each State should explain how, with a record of diplomatic exchanges that largely overlaps, it met this precondition but the other party did not. Both parties also insist that the breaches alleged target ethnicity and national origins (not nationality), to avert the outcome of Qatar v. UAE.

Both cases will likely proceed to the merits, in part at least. Surely, in each dispute, the respondent will argue that some allegations fall outside CERD’s substantive scope (for instance, the treatment of war prisoners and the laying of mines as such) or its temporal one. In this vein, Armenia described the Azerbaijani’s PM request as “replete with issues patently outside the parameters of the CERD” (CR 2021/25, Kirakosyan ¶ 2).

The real dispute objection

In addition, each State could make a more oblique argument: the real dispute cannot be before the Court (so ignore the CERD-one, it’s a pretext). For instance, both parties invoke a dispute on the conditions of prisoners’ detention under the Third Geneva Convention, outside the Court’s jurisdiction. Therefore, each State could caution the Court against hearing this matter, including its CERD-related details.

In pleadings, Azerbaijan questioned whether Armenia’s claims on war prisoners are genuinely CERD-based:

“[i]t is easy to lose track, however, during [Armenia’s counsel’s] submissions, that these are proceedings under the CERD” (CR 2021/23, Goldsmith ¶3).

More generally, Azerbaijan stated that:

“Armenia now seeks to repackage certain of its complaints regarding the Second Garabagh War that are entirely unrelated to racial discrimination as requests for provisional measures under CERD” (CR 2021/21, Mammadov ¶8).

This “repackaging” allusion hints to a “real dispute” objection: if the Court unwraps the packaging and re-characterises the dispute, it would find it lying mostly outside its competence. This objection leans less towards lack of jurisdiction than abuse of process (a cause of inadmissibility, but as elusive as the “holy Grail”).

Respondents have often attempted this objection recently, without success at the Court, and limited success before UNCLOS tribunals; the scholarship took notice. Conceptually, it is a blurry objection. It might include, but does not correspond to, the “incidental issue” problem, arising when issues falling outside the Court’s competence need determination (i.e., are implicated) en route to determining the main claim. Insofar as it points to a claim’s context and background, the “real dispute objection” resembles rather the “political question” one, and is similarly toothless.

The “real dispute” routine usually goes as follows:

  1. the Court could hear the application, but should decline doing so due to its context;
  2. namely, a wider conflict exists, engaging other sources outside the Court’s competence;
  3. the wider conflict is the real dispute (in relative weight or importance);
  4. the Court should ignore the applicant’s characterisation, and accept that the entire conflict is outside its reach, including the fragment submitted before it.

This objection routinely fails, because it desperately points the finger at external issues: a dispute properly before the Court is accused of impropriety by association. Unsurprisingly, the Court has confidently replied (most recently, in 1955 Treaty and ICAO Art. 84) reciting a page from its “political question” playbook; the dispute’s context, including a wider conflict engaging other sources, cannot taint a valid jurisdictional basis. In the judgment (jurisdiction) in the 1955 Treaty case, the Court drew a visible link between the political question and the real dispute, effectively delisting wider conflicts, even legal ones, as political background noise (¶¶56, 59):

Certain acts may fall within the ambit of more than one instrument and a dispute relating to those acts may relate to the “interpretation or application” of more than one treaty or other instrument. … [however,] the Court cannot infer the subject-matter of a dispute from the political context in which the proceedings have been instituted, rather than basing itself on what the applicant has requested of it.

The rise of CERD litigation before the Court

The Court has received five CERD-based applications (besides these two, Georgia v. Russia; Ukraine v. Russia; Qatar v. UAE) in the past 13 years, four in the last four years. These applications emerged in scenarios of deep conflict between the litigants. Besides the Qatar v. UAE case, all the others disputes started in the wake of violence, with allegations of aggression, breaches of humanitarian law and international crimes. Both applications of September 2021 allude to genocide (ArA, ¶54; AzA, ¶¶10, 32).

When allegations span across so many other sources that seem more pertinent to the conduct, resort to the CERD has one plausible motive: it offers the only way to bring a dispute to the Court (short of the Genocide Convention). In turn, this pathway pre-determines the claims’ scope, which must be limited to CERD breaches. Jurisdictional convenience dictates the characterization of the alleged wrongdoing.

Armenia’s and Azerbaijan’s application would fit this format quite exactly, two caveats. First, both parties use this strategy, so they will hardly raise a dedicated objection. In this couplet of cases, each party will likely raise wide-ranging accusations, using the Court’s jurisdiction to inflict damage on each other and score some intermediate victories; i.e., continuing war by other means. Second, perhaps in this case the racial motive in the alleged wrongdoing is stronger, and it will be possible to bring inside the CERD-dispute some of the surrounding conflict.

The Court can determine responsibilities under the CERD, and perhaps offer some relief to the populations affected through protective interim measures. Yet, it is fair to predict that the conflict in and around Nagorno-Karabakh will not be resolved in The Hague (nor in Strasbourg). Like the Strasbourg tribunal, the Court “does not have the choice of which cases come before it”. Perhaps, States which genuinely long for resolution should refrain from burdening the Court with fragments of wider, intractable conflicts. The Court’s pronouncement on one artificially isolated aspect might entrench positions and hinder settlement.

First, the Court is not to blame for the multiplication of disputes using compromissory clauses like cookie-cutters. There is no established doctrine justifying turning them down, and if States indulge in this trend, the Court must oblige. These cases inevitably come with other motives than the peaceful settlement of the (real, wider) dispute(s). Tullio Treves succinctly defined them last week: “to expose to the Court and to public opinion egregious cases of human rights violations [or] to score points in a political controversy.” The Genocide cases (Bosnia v. Serbia and Croatia v. Serbia and the counterclaim) encapsulate this strategy at its purest: the applicants bet on the equation in the public eye between the legal fight (about a narrow crime) and the responsibility for the war at large. Arguably, the outcomes did not reward this calculation, nor settle the conflict between the parties.

Second, these ulterior motives do not affect an application’s jurisdictional propriety. However, using the Court to amplify entrenched recriminations might strain its legitimacy. Suffice it to note that both applications (and PM pleadings) abound in references, allusions, accusations and straight-on preteritions (CR 2021/25, Kirakosyan, ¶2) mentioning heinous crimes lying outside the borders of the CERD dispute. One wonders whether the States seised the Court to resolve or deepen the conflict.

Alpha disputes and the popularity of compromissory clauses

Alain Pellet warned:

“[i]t is not major and politically sensitive disputes that should be submitted to the Court, but the “lambda” [i.e., midway, average] disputes that poison bilateral relations [without threatening international peace and security].”

The conflict between Armenia and Azerbaijan has long undermined international peace and security: more alpha than lambda. There have been many like it.

Since 2000, the cases brought under compromissory clauses more than doubled those relying on unilateral declarations; the ratio is 5-to-1 in the past decade. Even controlling for Pact of Bogotá claims (brought under an open-ended clause), compromissory clauses remain the most invoked basis for the Court’s jurisdiction.

Yet, apart from two idiosyncratic examples, the latest treaty with a compromissory clause is the 2006 International Convention for the Protection of All Persons from Enforced Disappearance, entered into force in 2010. The most popular (CERD, VCDR, Chicago and Genocide Conventions) are ancient. The drought of new compromissory clauses is a fact, recorded by Tams in 2009 and the Rivista article in 2021. The decline of multilateral law-making certainly contributes to this drought, but it must be noted that compromissory clauses disappeared also from bilateral and plurilateral instruments, and even from the few new multilateral treaties (e.g., the Treaty on the Prohibition of Nuclear Weapons, entered into force on 22 January 2021).

A correlation might run between these two trends, given the aggressive use of compromissory clauses described, championed by Armenia and Azerbaijan. Visibly, applicants recruit the Court for lawfare, to prevail or recover ground in wider conflicts. Strategic use of compromissory clauses produces wrong-footed and irritated respondents, whose best defense is prevention: stop stipulating compromissory clauses. Compromissory clauses: popular with applicants, unpopular with drafters.


The CERD disputes between Armenia and Azerbaijan belong in a pattern of applicants trying to make jurisdictional ends meet. Actionable compromissory clauses are few and no State is keen to stipulate new ones, so would-be-litigants must squeeze old ones. CERD’s open-ended texture beckons the CERD-ification of disputes that are only incidentally about race discrimination, but feature inter-State animosity, and reference to the “equal enjoyment of human rights” facilitates cross-fertilisation with other sources. Animosity, of course, is common in international conflicts, so the arrival of new CERD disputes with immense stakes looming in the background is not surprising, and the States’ strategy is “blindingly obvious.” The CERD-angle, moreover, is still wider than the Genocide-angle, the closest available option.

The two pending cases have something novel: respondents might hesitate to mount a full “real dispute” objection, which could backfire in the converse dispute. Both parties appear largely unrestrained, and relatively uninterested in demanding restraint. The PM pleadings were symptomatic. Azerbaijan criticised the Armenian PM request for conveniently zooming-in on the 2020/21 events:

“It no doubt suits Armenia’s case to lift up only that small corner of the carpet: but lift up the carpet further and this case looks very different” (CR 2021/24, Lowe ¶2).

Armenia did not deny, and upped the ante:

“Azerbaijan wants to lift the carpet further. We can lift it completely off the floor” (CR 2021/25, Kirakosyan ¶5).

Armenia then listed massacres and pogroms the legality of which is not before the Court. There is so much dough left around the CERD-shaped hole, and the litigants crave to discuss it.

In short, both parties launched litigation to open an additional front of conflict; incidentally, the Court must resolve their CERD claims. The Armenian agent’s words, at once, explain the claim’s propriety ratione materiae and why the wider conflict’s resolution is not really a primary concern:

“Armenia does not ask that the Court resolve all of the many issues that have long divided the Parties to this case. Much less does it ask the Court to reverse all of the many consequences of the most recent armed conflict” (CR 2021/20, Kirakosyan ¶2).

These cases’ systemic effects are hard to discern, but statistics show two plausible trends. First, applicant States (and counsel) indulge increasingly in launching cases with ulterior motives. Second, States no longer include compromissory clauses in new treaties. The mirrored disputes of September 2021 permit drawing a plausible correlation. States behave opportunistically, but the opportunism of litigants and drafters is different. Litigants will be litigants, drafters beware.

Applicants squeeze any available compromissory clause to bring a lawsuit today; drafters make sure to minimise exposure to such lawsuits tomorrow. In between lies the Court. Its record as effective dispute-settler suffers each time it hears a dispute the solution of which cannot achieve a peaceful settlement in the real world. Collective self-restraint is not working, and the Court’s authority might wear out. Perhaps, truly, States should listen to Pellet’s warning and fill the Court’s docket with lambda cases.

Print Friendly, PDF & Email

Leave a Comment

Comments for this post are closed