I recently read a paper that got me thinking about the way ambiguity operates in legal texts, and in particular about what happens to that ambiguity when a text is translated into another language. Many of the points it raises are familiar from legal practice, but they take on an added significance in legal translation, where a choice of words can also become a choice between competing legal interpretations.
In Ambiguities in Normative Texts and Their Resolution in the Rule of Law[1], Pierre Moor argues that ambiguity is an unavoidable feature of legal systems. Legal rules are expressed through ordinary language, applied to facts that are often uncertain, and expected to govern situations that legislators cannot fully anticipate. Ambiguity therefore arises at several stages: in establishing what happened, in interpreting the legal text, and in deciding whether a particular set of facts falls within a legal concept. Courts are required to resolve these uncertainties and to justify the reasoning by which they do so.
The paper is a useful read for anyone interested in legal translation because it explains, from the perspective of legal theory, why difficult legal texts are difficult in the first place. For lawyers commissioning translations, including translations from Greek into English, many of the most challenging problems arise where language, legal concepts, factual circumstances and institutional practice interact. A translator may be able to identify a plausible English equivalent for every Greek word in a sentence and still face a serious question about what the provision means in law, how broad the relevant concept is, or whether the source text deliberately leaves room for more than one interpretation.
Moor’s discussion of legal language is particularly relevant here. He points out that legal norms have to be communicated through ordinary language and therefore inherit the ambiguities of ordinary language. Legal concepts are expressed through words, but their application depends on whether a particular real-world situation falls within the category represented by those words. His example of a “forest” makes the point clearly. Everyone has a general understanding of what a forest is, but a legal rule protecting forests immediately raises further questions when applied to a particular wooded area. Its size, ecological role, composition and place in the landscape may all become relevant to determining whether it is a “forest” for the purposes of the legislation. The difficulty lies in the legal scope of the category, and that difficulty remains present when the term is translated into another language.
This has an important consequence for legal translation. Ambiguity in a source text may be legally significant and may need to survive the translation. A translator who renders the provision in broader terms may produce fluent English, but would also have made a substantive interpretative choice about the scope of the rule. In litigation, arbitration, regulatory proceedings or contractual disputes, a choice of that kind may matter considerably.
Moor’s idea of “normative density” is also useful for thinking about translation. Some legal concepts are relatively precise and lend themselves more readily to deductive application. Others, including concepts such as “fault”, “just cause” and “public interest”, require an assessment of the circumstances and cannot be applied mechanically. The lower the normative density of a concept, the more room there is for judgment about whether the facts fall within it. The same distinction affects translation. A tightly defined procedural term may have a relatively stable rendering in English, whereas an open-ended concept may depend heavily on the statutory context, the relevant case law and the legal culture in which it operates. This is particularly important where the source and target languages belong to different legal traditions.
Greek-to-English legal translation illustrates the problem well. English terminology carries its own legal associations, and an apparently obvious English equivalent may suggest doctrines, procedures or institutional arrangements that are different from those of Greek law. The issue becomes especially acute with terms that have familiar-looking English counterparts. A translation may read smoothly and convincingly while leading an English-speaking lawyer towards a concept derived from English law rather than the concept operating in the Greek source. Careful translation therefore requires attention to the legal function of a term within its source system, the purpose of the document and the expectations of the intended reader. Depending on the context, the best solution may be a conventional functional equivalent, a more literal rendering that avoids importing unwanted associations, or an explanatory formulation.
The paper also helps to clarify the relationship between translation and interpretation. Moor’s judges are constantly engaged in interpretation because legal texts do not always determine their own application. When ambiguity exists, the judge must choose between possible readings and provide reasons for that choice. Moor distinguishes between arguments drawn from within the legal system and arguments that draw on wider social knowledge, political culture, experience and ideas of fairness. Translators also make interpretative decisions, although their institutional role is very different. A judge has authority to resolve a legal ambiguity. A translator generally does not. For that reason, where several interpretations of a source expression are genuinely possible, considerable care is required before the translation silently commits the reader to one of them. Preserving the ambiguity, choosing terminology that leaves the question open, or alerting the instructing lawyer to the issue may be preferable, depending on the purpose of the translation.
This is also where current enthusiasm for AI in legal translation needs to be treated with caution. AI may be useful for generating an initial draft, suggesting terminology, comparing possible formulations or identifying passages that deserve closer attention. It does not remove the interpretative problems described by Moor. A language model may produce a seemingly highly fluent English version of a difficult legal passage while obscuring the fact that the source admits more than one reading. It may choose the most statistically or linguistically plausible interpretation without recognising that the ambiguity itself carries legal significance. It can also select an English legal term that appears semantically close while importing concepts associated with a different legal system. In that sense, fluency can make the problem harder to detect, because the resulting text may appear authoritative and settled even where the source is neither.
AI also cannot supply the institutional judgment that Moor describes. His account of legal reasoning depends on context, legal authority, argumentation and an understanding of how a particular legal system treats an open-textured concept. An LLM may help retrieve or organise relevant information, but the decision that a Greek term should be translated in one way rather than another may ultimately depend on the purpose of the translation, the procedural setting, the surrounding legislation, relevant case law and the consequences of choosing one interpretation over another. These are questions that require legal and translational judgement. The greater the ambiguity of the source and the greater the legal consequences of resolving it, the less sensible it is to treat machine-generated fluency as evidence that the underlying problem has been solved.
Moor’s broader conclusion is that ambiguity has a legitimate place within law because it allows legal rules to remain responsive to circumstances that could not have been fully anticipated when they were drafted. That conclusion has an important implication for anyone commissioning a legal translation. Quality cannot be assessed only by the smoothness of the English or by whether each source-language term appears to have received a neat equivalent. A good legal translation must also preserve the distinctions, uncertainties and interpretative possibilities that matter in the source legal system. In difficult cases, some of the awkwardness belongs to the law itself.
For lawyers dealing with complex cross-border matters, Moor’s paper is therefore useful well beyond its immediate subject. It provides a theoretical explanation for something experienced legal translators encounter in practice every day: legal meaning does not sit neatly inside individual words waiting to be transferred from one language to another. It emerges from the relationship between text, legal concepts, facts, institutions and interpretation. Any translation process, whether human or AI-assisted, that overlooks those relationships risks producing English that is perfectly readable and legally wrong.
[1] Moor, P. (2026). Ambiguities in Normative Texts and Their Resolution in the Rule of Law. In: Wagner, A. (eds) International Handbook of Legal Language and Communication. Springer, Cham. https://doi.org/10.1007/978-3-031-87993-7_89-1
