Login | October 18, 2024
Pennsylvania bars weigh in on legal AI
RICHARD WEINER
Technology for Lawyers
Published: July 12, 2024
So legal generative AI is apparently, and unfortunately, a thing now and I have to continue to write about it. Not necessarily what I wanted to write about after five weeks in Italy, but here we are.
Our next door neighbors in Pennsylvania have produced some pretty good guidelines on the potential use of generative AI in writing up motions and other legal docs. That is, if you have to try Chat GPT or the other ones.
Really, don’t. If you can’t do your own research, then hire somebody who can.
But, if you’re gonna do it, here are some guardrails that make sense, courtesy of our Paennsylvania sisters and brothers.
1. Be truthful and accurate. Basically, use your own logic and reasoning in applying the law to the case. Don’t use Chat GPT’s legal reasoning.
2. Verify all citations. A primary weakness of using gen AO in legal research is pretty well known—it hallucinates cases, citations and case law. I don’t think that’s been cured yet. So double-check all cites.
3. Ensure competence in using the AI. Read about it, particularly its weaknesses, and plan to overcome those weaknesses. So, again, the time you spend doing this is probably better used in doing traditional research.
4. Maintain confidentiality. Anything typed into any gen AI model is then open to the public. Don’t put anything into the prompts that you don’t want everyone to have access to. Another major weakness in the legal AI arena.
5. Identify conflicts of interest. As per usual, although I’m not sure exactly how this is different from any other instances.
6. Communicate with clients. You had really better tell them you’re using AI and have them sign off on that use. And be able to explain what you’re doing with it.
7. Ensure that the AI is ethically trained and that its output conforms with legal ethics. Sorry, but unless you know who is training it and how that’s being done, this is literally impossible. You don’t know how it’s trained and by whom, and they won’t tell you. Oops.
The rest of it is about ethics—bill correctly, use professional judgment, etc.
Really-- stay away from legal Gen AI unless you really know what you’re doing with it. And you probably don’t. But if you do, just use it for writing the first draft of client letters. It’ll fail you for anything more complex.
Maybe the next generation of legal AI will be better than the current platforms.