Every legal output must be explainable, grounded in validated authority, testable and versioned.

The Four Laws of System Design for Computational Law state these requirements as an ordered set, and a system that violates any of them is defective as a matter of design. The order matters: as with Asimov's laws, the higher law controls when two conflict.

The Zeroth Law comes first: no output may be presented with greater confidence than the system can defend, and a refusal is always preferable to a fabrication. The First Law requires explainability: every result must be traceable, on demand, to its inputs, its reasoning and its supporting authorities, because lawyers rightly extend no default trust to a machine, and trust is earned by reasoning that can be reconstructed. The Second Law requires grounding: every legal proposition must rest on validated authority rather than on a model's training data or its reconstruction of doctrine. The Third Law requires that reasoning components be independently testable and version-controlled as first-class assets, so that each can be verified in isolation.

Why these four, and why as law rather than aspiration? Because legal work cannot accept the graceful degradation tolerated in other domains. A slightly wrong recommendation elsewhere costs little. A slightly wrong proposition, filed and cited, becomes part of a court's record and a client's outcome.

The laws are also testable against real systems. The isResponsive methodology for technology-assisted review ties each law to an enforceable control: hash-locked rubrics and stored artifacts for traceability, validation gates that refuse to certify completion on a classifier's say-so, and services with defined boundaries that can be tested one at a time. A law that maps to a control can be audited, which is the point of stating it as law.