There is a version of this conversation that happens often in legal and business contexts. Someone has a French contract, a set of board minutes, or a regulatory filing. They run it through DeepL or a similar tool, get a readable English output, and decide it is good enough.
Sometimes it is. For routine correspondence, internal summaries, or documents where approximate understanding is the goal, machine translation has become genuinely useful.
For French legal documents, the calculus is different.
The Two Legal Systems Problem
French law is based on the civil law tradition. US law is based on common law. These are not just different languages — they are different conceptual frameworks for thinking about obligations, liability, rights, and procedure.
The terminology that exists in one system does not always have a direct equivalent in the other. Machine translation tools handle this by choosing the closest-sounding English word. That approach works in general language. In legal documents, it can produce text that is linguistically fluent but legally incorrect.
Consider a few examples:
Gérant vs. manager vs. director. In French corporate law, a gérant is the manager of an SARL (a specific French company structure). Translating this as “manager” or “director” loses the legal specificity — and in a document that is being used to establish roles, responsibilities, or liability, that loss matters.
Mise en demeure. This is a formal notice with specific procedural weight under French law — it starts the clock on certain legal obligations. “Formal notice” is the common translation, but it does not carry the same connotations in English. A reader unfamiliar with French law may not understand the significance of what they are looking at.
Force majeure. Both legal systems use this term, but its application and the conditions it covers are not identical. A machine-translated contract may use the term correctly in a linguistic sense while obscuring a meaningful legal difference between the French and US versions of the concept.
What Machine Translation Optimizes For
Language models and neural translation tools are trained to produce fluent output. They optimize for readability and naturalness in the target language. This makes them excellent for general content.
They are not trained to identify legally significant distinctions, flag terms that don’t translate cleanly, or understand the context in which a document will be used. They will produce a confident, readable translation of a legal document regardless of whether that translation is legally accurate.
This is not a criticism of the technology. It is a description of what the technology is designed to do — and what it is not designed to do.
The Downstream Risk
Legal documents are used to do things: close deals, establish obligations, resolve disputes, satisfy regulators. A translation error in a contract does not become visible until someone acts on the document — and by then, the consequences may already be in motion.
The risks are not hypothetical. Poorly translated contracts have led to disputed obligations, failed deals, and litigation. Inaccurately translated regulatory filings have required resubmission, causing delays and additional expense. These outcomes are not common, but they are real, and they are disproportionately expensive relative to the cost of getting the translation right in the first place.
When Machine Translation Is Appropriate for Legal Contexts
This is not an argument against machine translation in general. It is a narrower point about high-stakes legal documents.
Machine translation is a reasonable starting point for:
- Initial review of a document to understand its general contents
- Internal summaries for preliminary due diligence
- Non-binding correspondence where approximate understanding is sufficient
- Documents where errors will be caught before the translation is acted upon
It is not appropriate as the final translation for:
- Contracts that will be signed and acted upon
- Regulatory or compliance filings
- Court or arbitration documents
- Documents that establish rights, obligations, or liability
The Standard for Legal Documents
Professional legal translation by a qualified human translator remains the appropriate standard for documents that will be used in legal or regulatory contexts. This means a translator who understands both legal systems, recognizes terms that require contextual judgment rather than dictionary equivalents, and will flag ambiguities in the source document rather than translating through them.
The cost difference between machine translation and professional translation is real. So is the cost difference between a clean transaction and one complicated by translation error.
J Translation provides French-English legal translation for businesses and professional services firms. Get in touch to discuss your document.