Insurance coverage counsel, take heart. In the most recent issue of his Coverage Opinions newsletter , Attorney Randy Maniloff published the results of an insurance coverage exam he recently administered to the ChatGPT AI engine. The results of the five question exam? “ChatGPt got 4 out of 5 wrong,” wrote Maniloff, “but not just wrong — dead wrong.”
ChatGPT, Esquire managed to answer only one question correctly concerning the non -insurability of punitive damages under Oklahoma law. The Bot botched the four remaining questions regarding aspects of the duty to defend, the pollution exclusion, and the insurer’s obligation to pay pre-tender defense costs.
Maniloff mercifully gave ChatGPT a bonus question which it also answered correctly. It is bad faith for an insurer to make a coverage determination using a Magic 8 ball, ChatGPT said. In three decades of practice, however, I’ve thankfully never been asked for a legal opinion on that particular question.
I caught up briefly with Randy after enjoying his piece, which like all of his writing, came with ample amounts of both wit and insight. “ChatGPT may – and that’s a may – have some value for general discussion of the law concerning a coverage issue,” Maniloff told me. “But coverage determinations are not made based on general discussions of legal issues,” he added.
The amusing exercise does offer a serious take-away. The analysis of insurance coverage questions requires highly specialized insight, and the ability to not only understand the import of policy provisions, but the nuance of the facts to which those provisions are applied. While AI might progress to some levels of further competence on the subject, it is not there currently.
Not today, ChatGPT. Not today…..
badfaithadvisor.com 