I sent a contact my first draft of an article about—what else!—dysfunctional mainstream contract language. In the modern way, my contact fed my draft to Claude, to see what Claude thought of it. They then forwarded to me what Claude had to say.
One thing stuck out:
Litigated language has known meaning; novel clearer language trades interpretive certainty for readability, and imposes negotiation friction on the other side’s review.
So Claude buys into the fallacy that mainstream contract language has been “tested” and so is a safer bet than newfangled stuff! I’ve written about that in 2006, 2013, 2016, 2017, 2023, and 2025.
If you want to read more about the notion of “tested” contract language, I suggest you start with my 2006 post (and the article by Robert W. Benson that it cites, The End of Legalese: The Game is Over). Here’s what my 2006 post says about the main problem with this notion:
If contract language came to be tested, it’s because it failed to clearly state the intent of the parties. Why rely on language that resulted in litigation? Instead, express any given concept clearly, so you don’t have to gamble on case law breathing into it the desired meaning.
It’s particularly entertaining that Claude has bought into the notion of “tested” contract language. I’ve long said that if you rely on AI to supply you with contract language, it will likely give you the dysfunction of mainstream contract language, because that’s what it will have been fed. Well, consistent with that, here we have Claude regurgitating a weak defense of the dysfunction.
If you’re aiming for something approximating mainstream contract language, AI might be just the thing for you. But that’s setting the bar low. Limbo-bar low.
What’s the alternative? There’s no easy option. As I discuss in this blog post, in a rational world, the starting point would be a library of highly customizable templates that comply with MSCD guidelines and were prepared with the input of specialists under strong editorial control.
