Our Take
Correct legal terminology masks incomplete legal meaning, especially across jurisdictions, because foundation models lack explicit mappings of where concepts diverge in scope and effect.
Why it matters
Cross-border legal work is already being shaped by this gap. As legal AI adoption spreads, fluent output that looks credible will become harder to audit, especially in routine workflows under deadline pressure.
Do this week
Legal teams: flag any cross-jurisdictional contract output to a human expert for concept-scope review before signing, focusing on enforcement conditions and remedies, not just terminology alignment.
Correct Words, Wrong Legal Framework
Legal AI systems are increasingly fluent. They produce outputs where definitions align with established usage, translations follow convention, and explanations reflect familiar legal structures. Michael Krallmann, CEO of TransLegal, identifies a critical blind spot: this linguistic and terminological correctness masks a deeper problem in cross-border work.
The example he cites is concrete. Liquidated damages in common law jurisdictions and contractual penalty clauses in civil law systems appear terminologically equivalent. AI translation systems map one to the other without hesitation, producing clean, professionally sound text. But the legal meaning diverges sharply. Common law enforces liquidated damages only if they represent a genuine pre-estimate of loss. Many civil law systems enforce penalty clauses even where they exceed compensatory loss, subject in some cases to judicial adjustment. Same label. Different legal consequences.
The problem isn't translation error alone. It's that when a lawyer sees a familiar term, embedded assumptions about enforceability, remedies, procedure, and interpretation activate automatically. Readers fill gaps without realizing it. Under time pressure or in routine workflows, outputs that read credible and precise can pass review without deeper scrutiny.
Data Architecture, Not Presentation
Foundation models are trained on large volumes of legal text, but not on structured representations of how legal concepts relate across jurisdictions. They lack explicit mappings of where concepts overlap, partially align, or lead to different outcomes. Faced with a gap, the model produces the most plausible approximation.
Prompting, interface design, and retrieval can improve presentation and relevance. They cannot solve this problem. If a system contains no information about how two concepts differ in scope or effect, it has no basis on which to flag those differences. Output remains fluent. The underlying legal position is incomplete.
This is not a problem that better review processes alone can address. The source sits in the data. Legal meaning needs to be represented explicitly, with attention to purpose, scope, and legal effect, and with clear identification of where concepts don't align cleanly. Such information doesn't emerge automatically from large corpora. It must be constructed, curated, and maintained.
What Cross-Border Teams Should Audit Now
As legal AI adoption expands into cross-border contracts, the quality of terminology will matter less than what sits behind it. Outputs can look entirely correct and still lead users in the wrong direction.
For practitioners: when reviewing AI-generated legal output for cross-jurisdictional work, do not treat terminological alignment as confirmation of legal accuracy. Audit the scope and enforceability conditions specific to each jurisdiction. Verify that remedies, procedural context, and interpretation rules match your target jurisdiction, not just the labels. Under deadline pressure, this is where credible-looking errors slip through.