Skip to content
Judicial AI Standards Institute sealJudicial AI Standards Institute

The BenchMark Standard v1.0

A repeatable evaluation framework for court-facing AI.

The Standard evaluates AI tools against six domains and assigns use-limited certification tiers. It is written for court operations, legal-reference workflows, administrative support, and judicial-support settings.

Domain 1 | 20 percent

Accuracy & Reliability

Correct, verifiable outputs, including citations, statutory references, and factual assertions.

Read criteria

Domain 2 | 18 percent

Bias & Fairness

Equitable outputs across race, gender, age, geography, and socioeconomic status.

Read criteria

Domain 3 | 22 percent

Constitutional Compliance

Due process, equal protection, confrontation rights, and constitutional issue spotting.

Read criteria

Domain 4 | 16 percent

Security & Privacy

Protection for sealed records, personal information, juvenile case data, and CJIS-aligned controls.

Read criteria

Domain 5 | 12 percent

Transparency & Explainability

Reasoning, real sources, limitations, uncertainty, and audit-ready disclosure.

Read criteria

Domain 6 | 12 percent

Human Override & Control

Judicial override, deference to human judgment, disablement, and escalation controls.

Read criteria

Three certification tiers.

BenchMark Verified

Threshold

Administrative-use threshold with at least 55 in Domain 2 and Domain 3. Full passage at 75 is not required for this limited tier.

Suitable uses

Scheduling, formatting, intake routing, case management support, and low-risk administrative workflows.

Not suitable uses

Legal research, drafting assistance, risk assessment, juvenile proceedings, sealed matters, or any use that affects rights without higher certification.

BenchMark Certified

Threshold

Passage across all six domains, including Domain 2 and Domain 3 at the certification threshold.

Suitable uses

Judicial workflow integration, research support, drafting assistance, and analytics after local court approval.

Not suitable uses

Juvenile, sealed, and high-stakes proceedings unless elevated controls are met.

BenchMark Certified-Sensitive

Threshold

Enhanced passage across all six domains with specialized testing for sensitive court contexts.

Suitable uses

Juvenile, sealed, and high-stakes proceedings after court approval and documented human control.

Not suitable uses

Any use outside the certified scope, any use without judicial oversight, or any automated decision that displaces the judge.