Scaling legal AI across jurisdictions requires abstraction. Without standardisation, categorisation and simplification, systems would struggle to operate across large and heterogeneous bodies of legal information. At the same time, every abstraction narrows the range of distinctions a system can meaningfully recognise.

Consider fiduciary duties. Different jurisdictions may recognise related ideas while defining their sources, beneficiaries, scope and consequences differently. Mapping them into a common category makes comparison and interoperability possible, while potentially concealing distinctions that matter in practice.

This creates a strategic trade-off. A system designed around relatively aggressive abstraction may scale more easily because differences between datasets have been reduced. A system that preserves more jurisdiction-specific detail carries greater informational complexity. Which approach is appropriate depends partly on the intended use. A broad functional correspondence may be entirely sufficient for some applications, whereas small differences in legal effect may be precisely what the user needs to know in others.

Abstraction decisions should therefore not be viewed simply as technical housekeeping because they determine what a system will later be capable of distinguishing. Once concepts have been merged into a shared category and the original distinctions discarded, a downstream model cannot necessarily recreate the missing structure simply through better reasoning.

Cross-border legal AI will always require simplification. Legal systems are too complex and too different to represent every relationship exhaustively. The important questions concern the level of precision an application requires, which distinctions can safely disappear and which have consequences significant enough to preserve.

Scale and precision are both valuable, but they can pull in different directions. Recognising that trade-off early is an important part of designing legal data for cross-border AI.