Thoughts about AI by Theresa

Thoughts about AI by Theresa

I just attended a seminar put on by Filevine about AI. The speakers were:

I went into this with a  very open mind. My big picture takeaway is, this technology has not been harnessed or had boundaries applied. There are two types of AI that were discussed, largely: 1. chatgpt style that culls info from the internet at large and drafts something for you, and 2. AI technology that pulls only from the information you provide, ie a file in FV. 

My biggest takeaways re concerns:

AI doesn’t distinguish between what’s true on the internet and what’s not true. The source of information for your AI prompt is all of the information that’s available. Anyone can put anything on the internet. Like Wikipedia. 

There is a significant concern about bias. The historic bias against marginalized groups is part of what’s out there in the world. That will be incorporated into AI because it’s already out there. Then you start doubling down on bias and marginalization. 

There’s a significant concern about privacy in general. Any information you’ve put out there – address, phone, DOB, etc. can be culled by AI. 

The expectation is that the use of AI will affect malpractice (E&O) insurance. One carrier has submitted guidelines warning about using AI.  

Theft, fraud, stalking, hacking, terrorism are risks

Current AI: potential automated responses in gmail, yahoo emails. Suggestions from Amazon or other websites for other products. Chat bots.

“These are not decision-making tools.” “You can optimize for the outcome of your client. These tools can’t do that.”

How Filevine is using AI:

Demands: 48 hours, not final 

AI fields: use FV prompts to extract info based on documents in FV – “help you consume massive amounts of information”

                -depo questions

                - ICD10 codes

                -what is the drafter of RPD worried about

                -depo/depo summaries

                -objection logs

                -comes with page references so you can check

                -anything that can be OCR’d can be interpreted – not handwriting, diagrams, photos yet

 AI blocks: meant to accelerate composition of documents

AI Sidebar: conversation with your data, how do perform processes in Sidebar

Immigration AI: help with forms

My Impression:

Licensed legal work is required to be done by lawyers. All of this implicates our fiduciary duties, our duty of competence, our duty of confidentiality, etc.

In terms of briefing, demands, discovery, summaries – we have all of the resources available to do this and do it easily and quickly. My gut feeling is that it’s overwhelm that could lead to relying on AI as a short cut. The biggest risk to using AI is bypassing the thinking, issue spotting and planning that differentiates good lawyers from exceptional lawyers. This is the big danger spot I see in AI. FV’s AI can draft deposition questions, for example. But it can’t take into account what the needle moving issues are in my case, how I’m going to set up a particular witnesses, what case history I’m going to emphasize for leverage, whether I have all of the medical records, etc. That’s lawyer work. That’s what makes high value cases higher value cases. And a blunder there can impact trial, as we know.

Even using google, we have the option to evaluate the source before we click on the website link. With AI (outside of your file), there is no vetting of the source and no verification of the information whatsoever. That’s a huge concern when AI is used as a short cut for briefing. We have briefing templates in Westlaw for any CA motion/opposition imaginable, and it’s all correct and up to date. For other states, Westlaw has civil litigation guides or some other resource that gives you the rules of law and all you have to do is copy and paste and insert the facts of your case. It’s not difficult or time consuming at all. I’ve done this routinely in Iowa, Georgia, Florida, New Mexico, Washington, for example.

Where I can see value in case specific AI (within your file in FV, for example) is if the lawyer already knows the file and uses AI to create a document from a file that’s been uploaded into FV, for example, and then vets what’s in the document. The issue there is I’m not sure how much time you save since you have to vet the info regardless and make sure the document is what you wanted/needed it to be. 

Here are some other relevant articles and resources: