The International Law Commission, in the May-June 2026 session provisionally adopted Draft Conclusions 1-5 for Non-Legally Binding International Agreement (NLBIAs). The very title of the agenda has been a subject of debate since its introduction, with States (See sixth committee statements of India, France, South Africa, Israel, Ireland, etc) not agreeing to the term “agreement”. The struggle to settle upon a term reflects a deeper tension at the heart of international law: the elusive boundary between legal obligation and political commitment. This post contends that terminology is far from a semantic exercise and leads to different consequences depending on the usage.
Blurring the distinction between legal and political commitments unsettles the foundation of treaty law. The analytical instability of hard/soft binary, and the inadequacy of the treaty concept to capture the diversity of contemporary instruments, have both been extensively examined with the diagnosis broadly accepted even where the remedy remained contested. Yet none of that engagement anticipated the recursive difficulty that the ILC’s agenda now makes visible: what happens when the vocabulary to describe it is itself contested.
The Terminology
States do not select between “arrangement”, “agreement”, and “instrument” at random, each label is intentional, signalling an intent to not be bound while hoping to appear politically serious without legal exposure. These three terms specifically have been proposed by various states in the Sixth Committee.
Agreement
This is the term proposed by Special Rapporteur, Mathias Forteau. Under Article 2(1)(a) of VCLT, a “treaty” is defined by four cumulative elements: it must be (i) an international agreement, (ii) concluded between States, (iii) in written form, and (iv) governed by international law, whether embodied in a single instrument or in two or more related instruments, whatever its particular designation. The phrase “whatever its particular designation” was intended, as the ILC’s own commentaries to VCLT make clear, to ensure that substance governs over label.
The difficulty is that the word “agreement” is routinely applied to both categories creating exactly the inferential difficulty we are facing. For instance, the Ministry of External Affairs of India (MEA), in its Standard Operating Procedure on the conclusion of International Treaties in India, (SOP) treats agreements to be legally binding and synonymous to treaties. Yet it also refers to instruments across both categories loosely as “agreements/MoUs” throughout the document. India’s Indian Treaties database, maintained by the Legal and Treaties Division of the MEA lists treaties, agreements and MoUs within a single searchable repository without a systematic legal differentiation between them at the point of public disclosure. Thus, “agreement” functions not as a terminus of legal classification but as a floating signifier capable of carrying binding intent when the context demands it and shedding that intent whenever convenient. The Special Rapporteur also acknowledges in his second report, that several states have objected to the use of “agreement” in the title. It risks implying legal formality that the instruments under study deliberately avoid. The concerns expressed by States, coupled with the persistent uncertainty surrounding the legal and conceptual contours of the term “agreement”, suggest that it may not constitute the most appropriate terminology for the instrument under consideration.
Arrangement
The term “arrangement” is the most functionally transparent in diplomatic vocabulary. When a State uses this term, it signals that what follows is a structure of cooperation rather than a source of legal obligation. For example, the Wassenaar Arrangement, wherein forty-two participating States coordinate export control policies under its auspices yet it is deliberately not a treaty and not binding.
As Arms Control Association reporting on the Arrangement confirms, even its “best practices” are non-legally binding. It does the work of framing multilateral coordination without submitting participating States to the treaty registration requirements of Article 102 of the UN Charter, or the interpretative machinery of the VCLT.
This understanding is further reflected in State practice. Australia expressly distinguishes terms such as “arrangements” as less than treaty status, from agreements which denote a treaty. This linguistic clarity makes “arrangement” diplomatically useful, allowing States to signal an intention to cooperate without assuming legal obligations. Yet the same clarity may also discourage closer scrutiny of the instrument’s legal character by signalling non-bindingness at the outset. It risks pre-empting the inquiry into intention, form and substance as required under Article 2(1)(a) of the VCLT. The use of a lesser label does not, by itself determine an instrument’s legal status. The terminology may therefore suggest that political question is settled before the legal inquiry has properly begun.
Instrument
“Instrument” is the broadest and indeterminate of the three. It functions as a genus label; a word capacious enough to cover treaties, MoUs, declarations, final acts etc without committing the State to any view on the legal character of what is being described.
The ICJ’s judgement of 19 May 2025 in Land and Maritime Delimitation and Sovereignty over Islands (Gabon/Equatorial Guinea), makes this function unusually visible. The Court was asked to decide whether the “Bata Convention” a 1974 document asserted by Gabon to constitute a binding legal title over their common boundaries, had the force of law. Its threshold question was whether the – “intention of the parties to be legally bound” could be established, the condition the Court treats as necessary for an instrument to constitute a treaty. To answer it, the Court approached the document first as an “instrument”, the neutral term and only then assessed for whether it satisfied the legal requirements for a treaty, weighing the terms of the document itself, the circumstances in which it was drawn up and the subsequent conduct of the Parties. Court concluded, by fourteen votes to one, that it was not a treaty having force of law. The majority found that although the document bore several features which, at first sight made it appear to be a treaty, certain features (Article 7 of Bata Convention and nota bene clause appended to Article 4 anticipating renegotiation to align it with 1900 convention) cast doubt on the Parties’ intention to definitively establish their common boundary. Judge ad-hoc Pinto, applying the same test to the same evidence, reached the opposite conclusion: Equatorial Guinea never alleged forgery, never disputed the Convention’s 1974 transmission to France, nor challenged the political context of its signing. Yet the majority read this as showing no intention to be bound.
The term “instrument” is the most defensible of the three, precisely because of the trade-offs the preceding two sections expose. “Arrangement” signals non-bindingness before the legal analysis has begun, foreclosing an inquiry Article 2(1)(a) requires to be kept open. “Agreement” carries binding connotations which causes confusion amongst states. “Instrument” avoids both errors by naming the document without pre-deciding its nature, deferring the entire classification question rather than answering it in advance.
That deferral comes at a cost. “Instrument” settles nothing, the entire burden of classification shifts to whoever applies Article 2(1)(a) later, usually a court weighing intention, form and conduct after the fact. It doesn’t make the weighing exercise any more certain, it only postpones it. What uncertainty can look like in practice is visible in the Gabon case.
The Two Pressure Points
There are two moments at which the absence of a stable vocabulary starts producing genuine legal failure.
- Judicial Classification
When a court is asked to determine whether a document is legally binding, it begins where every interpreter must; with the text. The label is often the most immediate signal available, and yet the case law has repeatedly demonstrated that it is almost never decisive. Article 2(1) of the VCLT, severed the link between designation and legal status by providing that an instrument qualifies as a treaty based on substance and intent, “whatever its particular designation”. Such a definition was premised on the assumption of stable categories of treaty and non-treaty. The proliferation of deliberately ambiguous labels has eroded that premise. A court reaching for the neutral term “instrument” to begin its analysis (as ICJ in Gabon/Equatorial Guinea) is performing a classificatory act before classification has begun. It is already choosing a term that prejudges nothing, precisely because the labels the parties chose have left everything open. If the ILC’s text settles on “agreement” as the term, courts inheriting that vocabulary may face a rebuttable inference of binding intent whenever the word appears in a disputed document, the opposite of what the Commission intends to produce.
- Normative Pull
The most consequential pressure point is normative pull, the capacity of an instrument to generate compliance that the parties did not accept. The Paris Agreement of 2015, is the perfect illustration because its drafters were conscious of the risk and engineered around it with care. This shows that the terminology problem this piece traces in titles and labels is not confined to them: the same normative pull can be generated by drafting register even where the instrument’s own designation is not in question. Article 4(1), which establishes the nationally determined contribution framework is drafted in the language of obligation: each Party “shall prepare, communicate and maintain” successive NDCs and “shall pursue domestic mitigation measures”. The syntax is indistinguishable from a binding treaty provision, the legal status is not. The procedural obligation to prepare, communicate and maintain NDCs is itself legally binding; what is not binding is the content of any given NDCs. A State may set whatever target it chooses and is under no obligation to achieve it. Yet the language did something the formal classification did not intend. The result was that NDCs began functioning in domestic and international discourse as though they carried the normative weight of the procedural obligation. Thus, drafted with sufficient formality, language generates compliance expectations before the legal analysis even begins. ILC’s choice of term for its own definitional text carries this same risk and burden.
Conclusion
The terminological debate surrounding NLBIAs is one about where law draws its foundational boundaries. Each of the terms under consideration carries consequences that extend beyond the document it labels into the judicial, constitutional and normative architecture that surrounds it. “Agreement” risks generating binding inferences the Commission explicitly intends to avoid. “Arrangement” forecloses legal analysis before it begins by signalling non-bindingness at the threshold. “Instrument,” names the documentary vehicle without pre-determining its legal character and preserves the analytical openness the Commission’s project requires. The Commission should adopt “instrument” as the operative term in its final conclusions because it is the only one that does not answer, before the law has spoken. A neutrality the ICJ itself relied on when it approached the Bata Convention as an “instrument” before assessing whether it satisfied the requirements of a treaty.
Rhuhi Dhruvpal Sisodia, a fourth-year law BBA-LLB law student at Gujarat National Law University
Image Credit: International Law Commission, Seventy-seventh Session (2026), United Nations.
Jindal Forum for International and Economic Laws 