many lawyers report these products aren’t reliable enough for them to trust

First, the lawyers will learn from experience that current probabilistic AI is inherently unreliable. And having learned this, they will then direct their attention to others who adopt and make use of AI for any sort of "serious" work.

Applying technology that is widely known to be unreliable fits the classic legal definition of "negligence".