As the New York Times reported in December, “ChatGPT is, quite simply, the best artificial intelligence chatbot ever released to the general public.” Built by OpenAI, a San Francisco-based company, ChatGPT has grabbed headlines over the last two months.
Artificial intelligence, including chatbots, has myriad applications in the practice of law. AI no doubt will generate extensive online content readily visible when lawyers and litigants conduct quick online research using search engines like Google. How good, then, is the current version of ChatGPT? I asked the chatbot to write a post about undue influence, a common issue in California trust and estate litigation and the subject of many Trust on Trial posts.
I went to ChatGPT’s website (currently available at no charge) and typed: “Write a blog post for lawyers about undue influence claims in California trust and estate disputes.” The answer came quickly:
Undue influence is a legal claim that can be made in trust and estate disputes in California. It refers to the situation where one person exerts excessive control or persuasion over another person, causing them to act in a way that is not in their own best interest.
In trust and estate disputes, undue influence claims are often made in relation to the creation or amendment of a trust or will. For example, a person may claim that a parent or other relative exerted undue influence over them in order to change the terms of a trust or will to their benefit.
To prove undue influence in California, the person making the claim must show that the influencer had a fiduciary relationship with the person they influenced, that the influencer had the opportunity to exert undue influence, and that the influencer actually did exert undue influence.
It is important to note that undue influence claims can be difficult to prove, as they often rely on circumstantial evidence. Therefore, it is important for lawyers handling these cases to gather as much evidence as possible, such as witness statements and financial records, to support their client’s claim.
If a court finds that undue influence occurred, the trust or will in question may be set aside or amended to restore the intended distribution of assets. In addition, the influencer may be held liable for any damages caused by their actions.
It is also important for lawyers to be aware that in California there are statutory presumptions of undue influence in certain situations. For example, if a will is executed by a person who is not of sound mind, there is a presumption of undue influence.
How Accurate was ChatGPT?
ChatGPT’s sample post conveys some concepts correctly and reads well, perhaps lulling the reader to accept its accuracy. Yet there are major flaws in the analysis, leading me to grade ChatGPT’s post with a D.
Yes, under Welfare and Institutions Code section 15610.70, “excessive persuasion” is the touchstone of undue influence under California law. The real question, per the statute, is whether such persuasion “causes another person to act or refrain from acting by overcoming that person’s free will,” not (as ChatGPT states) whether the person is pushed to act against “their own best interest.”
To put this in context, Son may pressure Dad to disinherit Daughter in Dad’s trust, and such coercion may rise to the level of undue influence without reaching the murky question of Dad’s “best interest.”
ChatGPT teaches that a person claiming undue influence must prove the alleged wrongdoer had a fiduciary relationship with the victim. Wrong. While such a fiduciary relationship often exists and supports a finding of undue influence, no such relationship is required.
Yes, ChatGPT accurately explains that undue influence cases can be difficult to prove and often arise from circumstantial evidence.
As to who bears the burden of proof, ChatGPT is correct that there are statutory presumptions of undue influence in some situations, but the one example ChatGPT offers misses the target. If a will is signed by a person who is not of sound mind (as determined under Probate Code sections 810, 811 and 6100.5), the court should invalidate the will and need not reach the distinct issue of undue influence.
ChatGPT instead might have referenced Probate Code section 21380. This statute, among other things, presumptively invalidates gifts – by will, trust or otherwise – to paid in-home caregivers who fall within the definition of “care custodian.”
Although ChatGPT is much smarter than the Magic 8-Ball, a toy many of us have enjoyed, a lawyer who relies upon ChatGPT’s summary of California undue influence law runs a high risk of being led astray.
Improvements in AI should close the accuracy gap. Perhaps ChatGPT already has processed and learned from this blog post. In any event, while chatbots may provide a starting point for analysis, lawyers for the foreseeable future should turn to more reliable summaries of California law.