Generative AI: What In-House Legal Departments Need to Know

The artificial intelligence (“AI”) products for lawyers released over the past several years have left a lot to be desired. Many were interesting but all lacked the “Oh, my God” AI moment. Until now. And unless you have been living on the moon for the past year or so, you know that ChatGPT has come to dominate technology headlines, not only in the business world generally but also in the world of legal services where, for perhaps the first time, people are starting to ask if lawyers can survive this technological tsunami . Below we discuss generative AI and how in-house legal departments will be affected.

What is artificial intelligence?

The term “artificial intelligence” can be a bit misleading, at least when it comes to application in the legal field. Artificial intelligence is an umbrella term to describe technologies that rely on data to make decisions. For purposes of the legal work, a better description is “cognitive computing.” Cognitive computing uses AI systems that simulate human thought to solve problems using neural networks and other technology. Cognitive tools are trained vs. programmed - learning how to complete tasks traditionally done by people. I like to think about it as a research assistant who can sift through the dreck and tell you what it found. Why is this important? Because 328.77 million terabytes of data are created each day. The ability of any human to review and comprehend that level of dreck is impossible. AI systems augment our ability to digest such a vast amount of data and generative AI adds a powerful method of doing just that.

What’s going on and why now?

In 1965, Gordon Moore made a prediction based on his observation that the number of transistors per square inch on integrated circuits had doubled every year since their invention. This is, of course, “Moore’s Law.” His law predicts that computer power will double roughly every two years while the cost of that computing power goes down. Simply put, more computer power for less money. When coupled with the ever-lower cost of data storage, you have the foundation for the rapid rise in AI capabilities and availability. For ChatGPT, the sudden explosion into our consciousness was caused by decades of scientific work that was finally matched with the right level of raw computer processing power to make it feasible to launch and use such technology, i.e., in late 2022 the technology caught up with the thinking.

As business leaders (and businesses) become adept at using these tools, they will expect the other members of the C-Suite – including the general counsel and the legal department – to follow suit. Legal departments, therefore, need to be ready for this change and must adapt quickly to the use of generative AI.

How it works

For our purposes, artificial intelligence has evolved in three stages, and understanding this evolution is key to understanding the power of generative AI:

Stage 1 - AI is a computer programmed to mimic human intelligence, e.g., it can recommend a song you might like, spot spam emails and move them out of your inbox, and even drive a car.

Stage 2 – On top of general AI came machine learning, a branch of artificial intelligence that allows a computer to learn from data without being specifically programmed. It's like teaching a computer to play chess. At first, the computer doesn't know how to play, but it gets better over time as it gets more and more information or experience.

Stage 3 – Now comes generative AI, which is like Picasso in the world of artificial intelligence. Instead of just learning patterns and making decisions like other versions of AI, it can create new stuff. It can write songs, paint pictures, design graphics, or even write stories.

Where lawyers once used AI to extract pertinent information by typing a query directly into the machine, lawyers can now ask it to create things. Because of the power of generative AI to create vs. regurgitate and to interact with the user in ways that mimic human behavior, lawyers are far more ready to adopt and use these tools than they were five or six years ago when AI first came on the legal scene. The result? Generative AI will become ubiquitous – an indispensable assistant to practically every lawyer.

Robot lawyer army?

Now I tackle the question on every lawyer’s mind – will generative AI replace lawyers? In short, I am sorry to disappoint anyone who had visions of unleashing a horde of mechanical robot lawyers to lay waste to their enemies via a mindless rampage of bone-chilling logic and robo-litigation. That isn’t happening (but what an HBO mini-series it would make!). Instead, what is likely to happen are three things.

1) Certain legal roles may undergo changes, particularly those primarily focused on tasks such as document review, summarization, or initial legal research, which could see shifts in their responsibilities.

2) Jobs will be created, including managing and developing generative AI (legal engineers), and writing prompts for AI (prompt engineering).

3) Most lawyers will be freed from certain mundane tasks and can focus more on work that creates value.

Moreover, the bar will not allow generative AI to replace lawyers; the practice of law will require humans in some capacity no matter what. Second, lawyers must validate everything generative AI spits out. Third, generative AI does not understand context, nor can it discern whether it is being used to come up with the answer the user wants vs. the correct legal answer (i.e., right vs. wrong, but also, as Charles Spurgeon said, right vs. almost right). Only people can do that. And fourth, when it comes to serious legal work, most clients will want to talk (or be able to talk) to a person, not a chatbot.

Generative AI should – eventually – make your life easier and allow legal departments to increase efficiency without adding (or cutting) headcount and without having to invest large sums of money. It will be another tool you can use to streamline tasks and reduce the amount of mundane work you must deal with.

The prompt

When it comes to generative AI, the most critical element is the “prompt” you use to get results. Prompts are questions, instructions, or requests that you type into the tool to trigger the process described above. You can ask generative AI multiple questions, make multiple requests as part of one prompt, or refine the results by refining the prompt or adding to the data generative AI is considering. It takes practice and time to get it right. The best way to think of prompts and generative AI is to treat the tool and process like you would treat a young associate or intern: You must brief it on what you want/need, give additional information over the course of the assignment, ask clarifying questions, and, in the end, likely fix the work product.

Generative AI works by attempting to understand your question (i.e., your “prompt”). Once you prompt it to do “something,” it scans a ton of information and writes the best answer possible based on the data it has access to (just like you do as a lawyer, only at a much slower pace). Here are some simple prompts I have created or found that make it easy to get started:

  • What is the standard for [set out legal issue] in [x] jurisdiction?
  • Outline the steps needed to do [y]
  • Can you explain the holding in [specific case]?
  • Create a checklist for…
  • Draft a contract for [scenario]
  • What are the ethical considerations for [specific legal scenario]?
  • Set out the pros and cons of [x]
  • Can you provide cases that discuss [specific legal issue]?
  • Summarize this agreement and identify the five most important terms

What can generative AI do right now?

Here is a short list of some of the things that generative AI can do for in-house lawyers right now:

  • E-Discovery document review and summaries
  • Legal research
  • Draft memoranda, contracts, clauses, email, legal briefs, and motions
  • M&A due diligence
  • Simplify text to the right level of understanding
  • Create checklists
  • Redline documents
  • Edit your writing

Let’s try it out!

Let’s say you just received the following request from the CEO: Can you come here in 15 minutes and walk me through how insider trading laws work? This is a perfect task for generative AI because you do not need to use confidential information when creating the prompt, and the law is fairly settled so you don’t need to worry too much about nuance in a general overview. Here is my prompt: Acting like a lawyer set out a basic overview of insider trader laws and regulations in the United States. And here is just part of the response I got back[WM12] in seconds:

Insider Trading Laws and Regulations in the United States: An Overview

Definition: Insider trading typically refers to the buying or selling of a security, in breach of a fiduciary duty or other relationship of trust and confidence, while in possession of material, nonpublic information about the security.

Primary Legislation: The primary laws governing insider trading in the U.S. are:

  • The Securities Act of 1933
  • The Securities Exchange Act of 1934, especially Section 10(b) and Rule 10b-5.

Material Nonpublic Information: This refers to information that a reasonable investor would consider important in making an investment decision and that hasn't been disclosed to the public.

Who Can Be Liable: Both individuals who trade on insider information and those who provide or "tip" the information to others can be held liable. This includes corporate officers, directors, employees, and even outsiders who may have access to confidential information.

Tippers and Tippees: A "tipper" is someone who discloses material nonpublic information to another person (a "tippee"). Both can be held liable if the tippee trades on that information.

Misappropriation Theory: This theory holds that a person commits fraud "in connection with" a securities transaction, and thereby violates Section 10(b) and Rule 10b-5, when he misappropriates confidential information for securities trading purposes, in breach of a duty owed to the source of the information.

Is it perfect? No. But is it good enough to allow me to brief the CEO in 15 minutes? Absolutely. And you can ask follow-up questions or turn the above into an email or a presentation.

Ethical dilemmas and AI

Let’s cut to the chase: generative AI has no ethics. It’s just a machine. Under the rules of professional responsibility, lawyers have many ethical obligations to comply with when using AI. Here are some of the most important under the ABA Model Rules:

  • Rule 1.1 – the duty of technical competence for lawyers (all lawyers must stay up to date on technological developments that impact the practice of law).
  • Rule 1.6 – the duty of confidentiality (when using generative AI, you must ensure that any client information you enter is not confidential or, if it is, that the tool you are using will protect that confidentiality).
  • Rule 5.3 – the duty to supervise non-lawyers (lawyers cannot outsource their work to non-lawyers, like ChatGPT. They must stay involved).

What should I do next?

Lawyers are slow to adopt new technology. We are naturally skeptical and see the problems with something new vs. the benefits. But ignoring generative AI is not an option. Here is what legal departments need to do next:

  • Embrace it. But act with restraint and caution.
  • Develop legal department and company policies regarding the use of generative AI.
  • Start small with free products and low-risk tasks (understanding the risks) and get your feet wet. Then move to the more powerful paid version. Finally, look for established companies offering generative AI products and use those to truly establish a foundation for AI use in the legal department, e.g., Thomson Reuters.
  • Do it as a team, i.e., figure out how best to make generative AI work for everyone in the department.
  • Keep data privacy and confidentiality concerns top of mind.
  • Learn how to draft prompts that work for in-house legal research and needs.
  • Understand your state’s ethical obligations around the use of generative AI.


Published .