AI Regulation Tracker  /  State law in force

Virginia Law Now Requires a Human Decision-Maker in Criminal Justice AI Decisions

Virginia Code section 19.2-11.14 has been in force since July 1, 2025. It requires that decisions across the criminal justice process be made by the responsible official, and it bars any such decision from being made without a human decision-maker. This is a binding state statute, and its reach is limited to criminal justice.

By Anthony Guerriero  ·  The Leveraged Years Regulatory Research  ·  Published July 16, 2026  ·  Last updated July 16, 2026

Virginia put a specific rule into its criminal code about where artificial intelligence can and cannot sit in a criminal justice decision. The section is short, and reading it in full is worth the two minutes. The heading is "Use of artificial intelligence-based tool," and it lives in Title 19.2, the part of the Code that governs criminal procedure.

The statute defines the technology in broad terms. An "artificial intelligence-based tool" means "any machine-based system or algorithm, including machine learning models, predictive analytics, and decision support systems, that analyze data and generate recommendations or predictions." That definition is wide enough to cover the pretrial risk-assessment scores, recidivism predictors, and other data-driven tools that have worked their way into courts and corrections over the past decade.

What the law actually requires

The core of the section is one paragraph. It says that "all decisions related to the pre-trial detention or release, prosecution, adjudication, sentencing, probation, parole, correctional supervision, or rehabilitation of criminal offenders shall be made by the judicial officer or other person charged with making such decision." Then it draws the line plainly: "No such decision shall be made without the involvement of a human decision-maker."

That is the whole point of the law. A tool can score, flag, or predict. It cannot decide. The named human who is responsible for the call, the judge, the magistrate, the prosecutor, the parole authority, has to be in the loop and has to own the outcome. You cannot hand the decision to the model and rubber-stamp what comes out.

The statute adds one more protection. "The use of any recommendation or prediction from an artificial intelligence-based tool shall be subject to any challenge or objection permitted by law." So if a risk score or a prediction is going to influence a bail decision or a sentence, the defense can contest it the way it could contest any other input. The output does not get to travel into a decision unexamined.

What this covers, and what it does not

I want to be careful here, because it is easy to stretch a law like this past what it says.

The reach is criminal justice, and only criminal justice. The list in the statute is the boundary: pre-trial detention or release, prosecution, adjudication, sentencing, probation, parole, correctional supervision, and rehabilitation. Those are government functions, carried out by public officials inside the justice system. The human decision-maker rule attaches to those functions.

It does not govern how a private company uses AI. It says nothing about hiring, firing, credit, insurance, housing, health care, or advertising. It does not create a general right to a human decision anywhere else in Virginia life, and it does not reach outside Virginia at all. If you run compliance for a bank or an employer, this specific section is not your rule. Virginia has separate legislative activity aimed at high-risk commercial AI, and that is a different track from this one.

It also does not ban the tools. Read carefully, the statute assumes that AI-based tools will be used and simply constrains how. A court or a corrections agency can still run a risk assessment. What it cannot do is let that assessment stand in for the human judgment the law reserves to a named official.

Why lawyers and compliance leaders should care

Even if you never practice in Virginia, this section is a useful marker. It is one of the first state statutes to say, in criminal procedure, that an automated system cannot be the sole basis for a consequential decision about a person's liberty. That framing, keep the tool, require the human, preserve the right to challenge the output, is the same shape you see forming in other high-risk AI proposals across states and sectors.

For defense and criminal practitioners in Virginia, it is a live tool. If a pretrial or sentencing recommendation leans on an AI-based score, the statute gives you a footing to ask how the tool works, to insist a human owns the decision, and to object to the output under existing rules.

For everyone else, treat it as a signal about where the line is being drawn. Regulators are not moving to prohibit these systems outright. They are moving to make sure a person stays accountable for the decision and that the machine's output can be examined. If you build or buy decision-support AI for any regulated function, that is the design posture to assume is coming: a documented human in the loop, and an output you can defend when someone challenges it.

What to do now

Read the section itself before you rely on any summary, including this one. If you work anywhere in Virginia's criminal justice process, confirm that a named official is recorded as the decision-maker on any determination that touches an AI-based recommendation, and that the recommendation is disclosed and open to challenge. If you are outside criminal justice, do not assume this statute gives you or your customers a right to a human decision, because it does not. And if you are designing AI decision-support for any high-stakes use, build for a human decision-maker and a challengeable output now, because that is the pattern the law is settling on.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any statute applies to your situation with qualified counsel in the relevant jurisdiction.