How Far Away from Practical Legal AI Are We?

How Far Away from Practic…

This article is far from the first and far from the best researched on the subject of AI applications in law. However, given that legal AI companies seem insistent on moving their products into the practice and the courtroom, and that there seems to be a huge motivation on their part to disrupt the legal business-scape, it seems worthwhile to examine how effective AI can be at tackling mankind’s most precise art and most inexact science.

We started by feeding a prompt to ChatGPT to get it to offer us some topics on which to write an article concerning Virginia law. It provided us several, though they were extremely broad. We picked the most specific one, which was updates and changes to Virginia’s criminal justice system including marijuana legalization, the end of mandatory minimums for certain criminal convictions, and seeking expungement for criminal convictions, along with a host of jail reforms like limiting solitary confinement. We asked it to write us a five paragraph article on the subject and this is what it put out:


“Virginia has recently passed several major criminal justice reforms, signaling a shift towards a more progressive and rehabilitative approach to criminal justice. These changes have been hailed as a major step forward in undoing the harm caused by the state's punitive criminal justice policies of the past.

One of the most significant changes is the legalization of marijuana, which went into effect on July 1, 2021. Virginia is now the first southern state to legalize marijuana for adult use, joining a growing number of states across the country. This move is expected to have a major impact on Virginia's criminal justice system, as it will reduce the number of people who are arrested and incarcerated for marijuana-related offenses.

Another major reform is the elimination of mandatory minimum sentences for some offenses. Under the previous system, judges were required to impose a minimum sentence for certain crimes, regardless of the circumstances of the case or the individual involved. This led to many people being sentenced to overly harsh and unjust sentences. The elimination of mandatory minimums is expected to give judges more discretion to impose fair and proportional sentences.

Virginia has also made it easier for people with criminal records to seek expungement of their records. Previously, only a limited number of offenses were eligible for expungement, and the process was complicated and expensive. Under the new law, more offenses are now eligible for expungement, and the process has been streamlined and made more affordable. This is expected to help many people with criminal records overcome the barriers to employment and housing that come with having a criminal record.

Finally, Virginia has enacted several other reforms aimed at reducing the harm caused by the criminal justice system. These include limiting the use of solitary confinement, improving conditions in jails and prisons, and expanding access to treatment and other services for people who are involved in the criminal justice system. These changes are part of a broader effort to create a criminal justice system that is more fair, humane, and effective at reducing crime and recidivism.

In conclusion, Virginia's recent criminal justice reforms are a major step forward in creating a more just and equitable society. By legalizing marijuana, eliminating mandatory minimums, and expanding access to expungement and other services, Virginia is taking a bold stance against mass incarceration and the failed policies of the past. While there is still much work to be done, these reforms represent a significant shift in the right direction and give hope to those who have been impacted by the criminal justice system.”


At first glance, you can see it isn’t bad, per se, but it isn’t quality legal scholarship, either. I’d rank it a High School level of scholarship because it is extremely broad and is quite noncommittal about its viewpoint. It refers to certain policies as “failed” without citing what those policies might be and in what way they have failed, it categorizes these reforms as a “major step forward in creating a more just and equitable society.” That’s an awful lot of editorializing for a chat program, and fails pretty spectacularly at how we talk about law. However, it did do something really special that we can tell – it pulled this out of thin air.

We didn’t feed it any of this research, it somehow aggregated data about recent events and produced this content. Lawyers rely on powerful algorithms every day to comb through centuries of legal precedents, case law, statutes, forms, briefs, administrative codes, and contracts to identify key terms and sort by like language in order to help us research. This is far more generalized than any of those tools, the program is able to identify what is significant regarding a broad set of criteria like “Virginia criminal justice reform.”

Currently, no AI is capable of arguing a case or examining a witness, but we would be quite interested in what an AI could do if exposed to a wealth of case law as well-curated as Lexis’ or Westlaw’s digital libraries. Rather than imagining an AI lawyer arguing a case before a jury, this evokes the image of a lawyer using a sort of AI paralegal who is able to generate real-time responses to unanticipated developments in a trial. It raises the possibility that no matter what occurs, powerful search, aggregation, and distillation of legal rules and principals, cases, statutes, etc., can be performed with astounding rapidity.

In the field of legal research and case management, it’s possible that this technology could reduce projects which currently take dozens or hundreds of hours by incredible percentages of that labor. Consider a mass torts case with over five thousand separate plaintiffs, each of whom must be responded to individually, but whose cases are extremely similar in nature, all having suffered the same injury. An AI advanced enough could take a lawyer’s response to one, or to several plaintiffs, and then extrapolate thorough responses for the remaining several thousand. Human lawyers would still need to check the work and edit where appropriate, but the time saved by preventing duplication of effort could dramatically reduce litigation costs.

We aren’t there yet, though. For now, AI is an interesting curiosity for lawyers and nothing more. But in the next ten years, given the pace of AI advancement, we may be looking at an extremely different landscape than the one we practice in today, one where our research and writing tasks are performed by tools and we become some sort of exceptions processors and editors, rather than authors and researchers ourselves.

If you have questions about this article, please contact Christopher Adams ( at 804-377-1273 or Steve Setliff ( at 804-377-1261.