AI & Law: Legally Defining AI

Image for post
Image for post
Considering various attempts at legally defining AI

by Dr. Lance B. Eliot

For a free podcast of this article, visit this link or find our AI & Law podcast series on Spotify, iTunes, iHeartRadio, plus on other audio services. For the latest trends about AI & Law, visit our website

Key briefing points about this article:

  • Definitions are the lifeblood of the law and inextricably the practice of law too
  • Many a court case has hinged precariously on the definition of this word or that phrase
  • Meanwhile, consider that AI is gradually and inexorably coming into the field of law
  • And yet there is ample ambiguity about what “AI” entails and definitions of AI are aplenty
  • This gives rise to examining a definition of AI as codified into the National Defense Act


Any budding attorney learns pretty quickly that definitions can make or break your day. Whether examining a recently passed piece of legislation or perhaps giving a contract some close inspection, the importance of definitions arises with near certainty. Is the terminology well-specified or is it sloppy and proffers numerous loopholes? Are the crucial words and phrases carefully defined or are they used with abandon and seem vacuous?

In contrast to the average person, lawyers know well the importance of definitions. Many a court case has hinged on the definition of a particular word or phrase. Tremendous legal battles have been waged as a result of arguments over the meaning of the tiniest of words.

Even the general public is likely aware of the now-classic debate over the word “is” when Bill Clinton during his presidency stated that “There’s nothing going on between us” and then explained to a grand jury that “It depends on what the meaning of the word ‘is’ is.” In this instance, the splitting of hairs dealt with whether the word “is” encapsulates the meaning of “was” or instead only meant the present tense and not the past.

Indubitably, definitions are crucial, and especially so in legal discourse.

Speaking of splitting hairs, another vital element about definitions entails what exactly the meaning of a “definition” itself entails. We all seem to accept the notion that a definition is that which defines something. This though is a bit recursive since it defines the word “definition” via the use of the word define, which might leave us meandering down a rabbit hole since we next need to stipulate the nature of the word “define” too.

Aristotle famously provided a definition of “definition” by indicating that it is a phrase that signifies a thing’s essence. Notice that this definition, fortunately, does not contain a self-referential denotation and therefore would seem to proffer some modest progress on clarifying the matter. Per his wise wording, we seem to vaguely know that a definition can be or is presumed to be a phrase, and the phrase provides a telling about a thing, for which the portrayal is intended to strike the core or essence about that thing.

Some linguists and philosophers do not especially agree with Aristotle’s indication. The angst involves the supposition that perhaps a definition is a kind of absolute. There is an implied sense that a definition is ironclad and abides by a law of nature such that it is inviolate and never changing. As such, the argument is that definitions are instead variable, ever-changing, and a social construct that merely reflects a type of crowdsourcing, namely that if enough people concur about a definition or even the definition of definition then that’s what it is.

In short, there are no absolutes, only relative meanings.

This brings us back to the legal viewpoint underlying all of this discussion. In theory, supposing that definitions are purely relative, it means that there is always going to be room for argument about the meaning of something. This then opens the door to enabling legal disputes, since you can nearly always find a path toward undercutting a definition or try to argue for an alternative definition that presumably supports your side and undermines the opposing side.

Legal argumentation has a lot to do with making arguments involving definitions. If a definition in a case seems to be detrimental to your legal posture, you can attempt to challenge the definition and fervently attack it. Meanwhile, if a definition seems to be supportive of your position, you’ll need to muster a strong defense to bolster and defend the definition.

A definition then is only as good as the strength of reasoning that keeps it intact.

There are ordinary definitions such as those found in the everyday dictionary and commonly used in ongoing societal affairs. For most lawyers, the definitions that really hit home are those that are statutory. For example, when a legislative body passes a law and includes various definitions, those become a kind of contract between the lawmakers and the people subject to those laws. The language specified is presumably laying a legal grounding and provides a semantic foundation for expressing what the law portends.

Definitions can be expressed by a concepts-oriented approach, depicting what a given word is supposed to mean, abstractly, and do so by offering a sentence or two to explain what it is all about. Another approach involves providing an example of what the word means. You can also try articulating the constituent parts or components that indicate the meaning. There is the approach of defining something by negation, stating what it is not, and therefore allow the reader to understand the word via knowing what it presumably must otherwise be. You can use enumeration. And so on.

This illuminates that there are multiple methods for composing or stating a definition, all of which have their own respective strengths and weaknesses, and for which a canny lawyer can leverage accordingly when disputing the definitions involved in a court case.

One would hope that a legislative body would be mindfully careful when providing definitions and attempt to make those definitions as bulletproof as possible. The more slippage or leeway within the definitions would suggest that the more ambiguous will be the results. How can people accede to what the law states if the law cannot deterministically and unambiguously clarify what the law presages?

Now that we’ve covered that muddy and rutty road about the maladies associated with definitions, let’s turn our attention to a definition of increasing need and qualm.

Defining AI Is No Piece Of Cake

What is the definition of Artificial Intelligence (AI)?

I’m guessing that most have not given the topic much attention per se. You daily are bound to hear or read the AI moniker, oft used in blaring headlines, and merely take at face value that it has something to do with intelligence and something to do with being artificial. Perhaps the imagery of a robot comes to mind, or possibly you hark back to HAL in the movie 2001: A Space Odyssey, and that is what your definition of AI embodies.

In a prior article, I discussed the meaning of AI as somewhat ascertained via the Turing Test, which involves a type of “I’ll know it when I see it” method to defining AI. That avenue is not particularly satisfying when it comes to passing laws and establishing contracts, thus other approaches must be assuredly undertaken.

For an attorney, knowing the definition of AI might seem nothing more than an idle activity and pretty much inconsequential.

I dare say that the meaning of AI will gradually and inexorably have a tremendous import for lawyers, for the practice of law, and the nature of our courts and justice.

Here’s why.

AI will be infused into the making of laws and the promulgation of laws. AI-based legal reasoning systems will partake in the performing of legal tasks, at times supplementing the acts of human lawyers and inevitably autonomously doing so. Ergo, being aware of and having an understanding of AI is tantamount to anticipating the future of being in the law and serving as an attorney.

Also, the chances are high that you’ll begin to see AI emerging amidst the legal cases that you deal with.

For example, an attorney specializing in real estate is certainly going to eventually find themselves immersed in a legal case that has at issue an AI-based real estate system that perhaps led someone astray or make “judgments” that are at question in the case. Overall, a modern-day attorney will ultimately see AI permeating all areas of legal life and therefore will have to recognize when such AI is at the heart of a case and how to contend with the AI injection thereof.

Yet another reason for a lawyer to be aware of the meaning of AI involves examining a newly passed law that explicitly identifies AI as a part of the legal wording of the regulation. Or, maybe when aiding the crafting of a law or perhaps drafting a private contract, there will be a need to incorporate the notion of AI and therefore its definition therein too.

If none of that yet seems compelling about why having a definition of AI is significant, perhaps some added fuel to the fire might help.

There is ongoing and rancorous debate about whether AI ought to be somehow be decreed as having personhood, gaining a semblance of rights and privileges akin to that of humans. It would seem any attorney of any specialty would want to partake in such discussions, given the magnitude of the legal repercussions. The same can be said for thorny legal issues such as the liability associated with AI systems, the responsibilities of AI systems, and so on.

Returning then to the question about what is AI, or more definitively what is the definition of AI, some trace the earliest establishment to the efforts of Professor John McCarthy and his indication in a 1955 paper that said this: “For the present purpose the artificial intelligence problem is taken to be that of making a machine behave in ways that would be called intelligent if a human were so behaving.” He also stated that AI is the matter of “getting a computer to do things which, when done by people, are said to involve intelligence.”

Years later, in a 2007 paper, he defined AI this way: “It is the science and engineering of making intelligent machines, especially intelligent computer programs.”

For details about the latest pace and trends of AI and the Law, including sample code excerpts and salient explanations, see my textbook entitled “AI and Legal Reasoning Essentials” at this link here:

Those initial definitions are akin to what you would see in most dictionaries today and have generally stood the test of time. But if you take a closer look, you would realize that such a definition is wanting in many sorrowful ways. The shortness, ambiguity, and semantic flimsiness is nearly overwhelming and makes those definitions seemingly inoperable for any suitable legal purposes.

Exemplar Of AI As Defined In The Law

We therefore can turn out attention to definitions of AI that have been codified into the law.

Let’s use the one such instance found in the National Defense Authorization Act of Fiscal Year 2019 as an exemplar.

Within the portion labeled as Division A: Department of Defense Authorizations, and within it, the sub-portion labeled Title II: Research, Development, Test, and Evaluation, and within that the Subtitle B: Program Requirements, Restrictions, and Limitations, you would find this definition of AI as indicated in Section 238 of the passed bill:

SEC. 238. Joint Artificial Intelligence Research, Development, And Transition Activities.

(g) ARTIFICIAL INTELLIGENCE DEFINED. In this section, the term ‘‘artificial intelligence’’ includes the following:

(1) Any artificial system that performs tasks under varying and unpredictable circumstances without significant human oversight, or that can learn from experience and improve performance when exposed to data sets.

(2) An artificial system developed in computer software, physical hardware, or other context that solves tasks requiring human-like perception, cognition, planning, learning, communication, or physical action.

(3) An artificial system designed to think or act like a human, including cognitive architectures and neural networks.

(4) A set of techniques, including machine learning, that is designed to approximate a cognitive task.

(5) An artificial system designed to act rationally, including an intelligent software agent or embodied robot that achieves goals using perception, planning, reasoning, learning, communicating, decision making, and acting.

This is by all appearances a much meatier definition of Artificial Intelligence.

The definitional approach focuses on a concepts-oriented indication, along with a type of enumeration that depicts several variants and presumably gains added definitiveness accordingly. The preface indicates the term Artificial Intelligence includes those five statements, which perhaps already raises some ambiguity as to whether they are all-encompassing or merely somehow representative, and likewise whether they are each mandatory or possibly separate and independent of each other (notably, the same definition appears later on in Section 1051 of the bill, repeating the same five statements and yet prefaces it with the indication that Artificial Intelligence “includes each of the following” and thus slightly differs in the precise wording therein).

In any case, this lengthier and legally opined definition of AI is nonetheless not necessarily overly tightened or shed of any openings or loopholes. There are numerous trapdoors and pitfalls encompassed, and I’ll provide a brief delectable tasting for you of especially noteworthy eye-openers.

The first statement indicates that AI is apparently a system that performs tasks “without significant human oversight” and therefore the definition appears to be striving toward the notion of incorporating autonomous operations (for more on the topic of defining AI autonomy, see my Daily Journal column of September 11, 2020).

On the one hand, yes, it is likely crucial to ensure that autonomy is noted as a key factor for defining AI, but in the same breath, the admission undercuts the premise by teasing with the ambiguity over the extent of human oversight. What does “without significant oversight” herald? Where does any oversight seemingly begin and end, such that we could all agree as to the degree of significant versus insignificant levels of oversight?

Envision a court case involving an alleged AI system. If the purported AI operated without human oversight for 90% of the time and required human oversight for 10% of the time, would that be sufficient to warrant being classified as an AI system under this first statement definition? What about such a system that was 99% of the time operating without human supervision and only 1% of the time with human oversight? You could readily make a compelling case that even in the lowly 1% instance, perhaps the oversight was trivial in the 99% of the time but life-determining in the 1% of the time, and thus you could argue that the tinier percentage is not a turning point per se.

Consider another quick attack at the language used.

The fifth statement indicates that AI is an “artificial system designed to act rationally,” implying that being rational is the essence of AI. Why is that so? This falls into the classic trap that robots and AI are somehow going to be purely logical and rational, as though they will never be subject to presumed human frailties and foibles. I’ve dispelled that myth in many of my writings and analyses. In addition to that flaw in the language, it also says that the system was “designed” to such action, which leaves wide open the difference between the design of the AI versus the resultant actions of the AI.

You can drive a Mack truck through that opening.

There are numerous other issues in the stated definition of AI, glaringly too its inclusion of “machine learning” as a seemingly vital ingredient, and yet nowhere else within the body of the bill is there a definition of Machine Learning (yet another term bandied around these days and for which there are an exceedingly large wide variation and heated disputable meanings).

All told, at least you’ve now had an opportunity to take a somewhat close look at how AI has been defined, readying you for the upcoming tsunamis of AI references and inclusions into our laws, into contracts, and into the daily lives of those practicing the law.


A rose might be a rose by any other name, but AI is not a rose and will increasingly be used in ways that have enormous legal repercussions, as such, we ought to be prepared for speaking and writing about AI in a manner that is sensible and usable. You can bet your bottom dollar that eventually you are going to be on one side of an AI definition, perhaps arguing it to the teeth as airtight, or you’ll be on the other side, clamoring that it is vacuous and inapplicable to your side of the case.

May the best argument win.

For the latest trends about AI & Law, visit our website

Additional writings by Dr. Lance Eliot:

And to follow Dr. Lance Eliot (@LanceEliot) on Twitter use:

Copyright © 2020 Dr. Lance Eliot. All Rights Reserved.

Written by

Dr. Lance B. Eliot is a renowned global expert on AI, Stanford Fellow at Stanford University, was a professor at USC, headed an AI Lab, top exec at a major VC.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store