Given the rapid expansion and seemingly endless uses of AI-generated writing tools, Crafty Counsel has been in touch with in-house lawyers in our community to ask them how they’re going about using programmes like ChatGPT and what advice they can give to those who are just dipping in the AI pool. 

Written by Laura Jeffords Greenberg, General Counsel at Worksome

How have you been using GPT-4? 

I’ve been using GPT-4 in a variety of ways. But I’m really excited about the recent release of  the browsing (with Bing) and plug-ins features. I’ve have been experimenting with these and they’ve improved the following use cases:  

💻 Basic Research:

ChatGPT is trained on a diverse range of internet text so it can be a good starting point for research on new topics. For example, I am a California lawyer working in Europe for a company that has operations in Australia. Our Australian banking partner had a question about whether a certain regulation would be applicable to our business. I had no idea where to start so I started with ChatGPT. I had a conversation with GPT-4, which provided a quick overview and lay of the land as it related to our business model and financial regulations in Australia. From there I was able to go to primary sources to craft an answer. It was much faster than initial searching on legal databases or Google searching. 

In addition, I’ve also used GPT-4 to understand why certain French legislation was excluded from a contract. It gave me a starting point to verify the information. After verification, GPT-4 helped me revise the clause and draft an explanatory comment for the redlined contract. 

Although ChatGPT has been trained with a mixture of licensed data, data created by human trainers, and publicly available data, it doesn’t have specific documents or sources in its training set. To be clear, it is not intended to be used as a primary source for legal research. ChatGPT can be used as a starting point for research but all output should be validated. 

💻 Improving Arguments and Positions: 

ChatGPT is a great tool for coming up with new ideas or perspectives that you may not have considered. For example, I use it to generate arguments to support my position in contract negotiations. I also use ChatGPT to generate arguments that could be put forward by the other side, which helps me address anticipated counter arguments before they are presented. 

You can do this by asking ChatGPT for both sides of an argument and you can specify whether you want arguments for or against a particular point. If ChatGPT gives you an argument that you find interesting but not fully developed, you can ask it to elaborate or explain further. You can also request different types of arguments. You can ask ChatGPT to come up with logical, emotional, or ethical arguments on a specific point. This can help you approach the negotiation from different angles.

💻 Understanding Concepts:

If you’re grappling with a complex concept or topic, ChatGPT can often provide clear, understandable explanations. It can also summarize and highlight the key points of complex documents or information. However, the validity of the explanation should be cross-referenced with reliable sources, as it can sometimes generate plausible-sounding but incorrect or misleading information. Rumors have become reality with the recent example of it creating non-existent cases in Mata v. Avianca, Inc. Always verify! 

💻 Writing and Communication Assistance:

ChatGPT can assist with a variety of writing tasks, including composing text, preparing outlines, and drafting sections of contracts or other documents. It can proofread for grammar and syntax errors, re-write text in plain English or with a different tone or style, and can check for consistency and clarity. 

Rather than me thinking, organizing and writing out various policy documents, emails or communication, ChatGPT can help me summarize and re-draft to create crisp communications – saving a tremendous amount of time. However, it’s important to review and edit the generated content to ensure it’s accurate and appropriate for your needs. 

My dream is to fine-tune an AI tool built on ChatGPT’s API so that the output is written in my voice and style – a personal AI (akin to Grimes’s AI project).

💻Creating Checklists for New Areas of Practice:

ChatGPT can help you prepare for new areas of practice, such as providing an overview or checklist of considerations. At Worksome we tested it out, asking ChatGPT to generate considerations in specific states when hiring an employee. It was able to pull information even at the city level. This can help ensure you’re prepared and have considered all relevant factors before approaching outside counsel, entering a discussion with the business or making decisions. For me, ChatGPT can help get me from 0% to 80-90% in a short amount of time at virtually no cost. This is a game changer for small legal teams. 

💻Creating Processes, Checklists and Templates:

Similar to creating a checklist or overview for new practice areas, I’ve created in-house GDPR compliance program using Chat GPT-3.5 through 25 prompts, created a template response to various types of emails such bug-bounty, and have showcased how to create a limitations to liability clause drafting prompt in an upcoming Contract Nerds “Chatting with GPT” article. 

Be aware of ChatGPT’s limitations

As mentioned above, it’s important to remember the limitations of ChatGPT. It generates responses based on patterns of human communication it learned during training and can sometimes produce incorrect or nonexistent sources. In addition ChatGPT only goes up to 2021 and has less information about esoteric things (but with the Pro subscription you can now browse the internet with Bing and use plugins (including a PDF and hyperlink reader)). Always verify! 

Lastly, as I am sure everyone is aware of now, when using ChatGPT with the Chat History & Training on, you are sharing your input with OpenAI. Do not include any personal data, confidential, privileged or third party information in the prompts. You can turn off the Chat History & Training or use ChatGPT’s API to try to mitigate such issues (although there are still concerns about whether this provides an additional layer of confidentiality and security). Because it’s so new, I use ChatGPT with the expectation that my prompts will become public. If I am concerned about the information I want to input, I sanitize it before inputting into ChatGPT.

What would be your advice for in-housers who want to start using GPT-4 but don’t know where to start?

My number one tip is learn by doing! Start experimenting with ChatGPT. Here are some useful resources to help you get started: 

What are you planning to do next? 

I want to start automating processes by leveraging ChatGPT’s ability to program/write code/develop apps or simple software. I want to challenge myself to see whether I can take a manual process and create a chatbot or app without assistance from an engineer. 

I’d also love to learn how to fine-tune a large language model to help it learn a certain style of writing and input company documentation so the output is customized for the company and is written in the appropriate company tone.

The pace of change and growth is difficult to keep up with so all of this may change tomorrow!