Hey, I get it. There’s no interest like self-interest, and when you feel you just might be becoming a relic… you do some silly things to try and stay relevant. But when that desperate attempt hurts students, it’s not okay. And there really is no other explanation for a Berkeley Law anti-AI policy that is… well, professor-protectionist if not downright speciesist.
Imagine you’d like to brainstorm a paper topic or thesis. It might be good to ‘kick the tires’ with the most knowledgeable systems on the planet, right? Well, not at Berkeley Law. That would be prohibited conceptualizing… which, yeah, sounds about as stupid as it is. It’s like banning the encyclopedia of yesteryear because, god forbid, you might actually learn something in the process. Instead, it would clearly be best to stumble around in the dark because, hey, it was uphill both ways when we went to law school… so, by god it is going to be that way for you as well! Why, we ought to get you Shepardizing in the books too, now that I think on it! <Sigh>
And don’t even get me started on using the best systems to improve your organization. Why, that would be prohibited outlining! And how could you, really?!
“Asking an AI tool to identify repetitive passages in a paper that should be cut”? Prohibited revising, for god’s sakes! What do you think this is? The twenty-first century? Why, in 1400, we didn’t even have a printing press, so surely you can make do!
What if you come upon a paper written in some language that you don’t speak? Well, enroll in a language class, of course! Please don’t tell me that your first instinct was prohibited translating! Young people these days… going after the best learning regardless of the human language in which it was written. Didn’t you read about the Tower of Babel? That was intentional, you know!
“Students may not upload course materials—including assignments, readings, slides, class recordings, or other class content—into generative AI systems.” But of course… protectionism.
Now, full disclosure: when I began teaching in 2001, yes, we did have discussions about this sort of thing. ‘Hey, I’ve worked hard on my course organization, and I’m the author of it, so it’s protected by copyright and you ought not post even your own outline to the internet!’ We absolutely did toss around such things. But that was 2001. We learned. It’s 2026, well past time to let it go. Information wants to be free… except at UC Berkeley Law, apparently.
Thank goodness “instructors have the discretion to deviate from this default rule.” If in writing. And with appropriate notice. And if they “require students to disclose any authorized AI use.” Goodness… I’m not sure I’ve ever seen folks try so hard to keep people from doing better work. Thou “doth protest too much, methinks.”
[Berkeley Notice: Google’s extractive AI was used to check the veracity of that memorized quote, and to see that it comes from Hamlet Act 3, Scene 2. Shame! Shame!… I know. Since Gemini chimed in, it also reminded me of the details of that original passage, and how the meaning of the phrase has changed over time. More shame! Such learning should have required at least one visit to the library, and maybe an office hour.]
ADDENDUM 5.28.26
A thoughtful colleague asked why I am against AI disclosure requirements – the reasons are several.
First, and this strongly informs my reaction to the entire Berkeley anti-AI policy, a teacher ought to always ask the following about every classroom policy: (1) does this, Satan-style, encourage cheating, and (2) does this advantage cheaters. Just as I am very much against policing operations that create criminals, I am very much against education policies that create cheaters. That is precisely what Berkeley’s policy will do. Every rational student will conclude (1) most students will ignore the inane policy and (2) their performance will be the better for it. Thus, those who comply will suffer grading consequences. That is reason enough to scrap the policy since its benefits are entirely imagined. Moreover, even as to ‘mere disclosure,’ the ‘honesty tax’ is real: it is impossible to know whether an idiosyncratic professor will consider particular AI use ‘overuse’ and penalize grading accordingly—or, even worse, trigger honor code proceedings. So, the rational thing is to understate—meaning to lie. Again, that is reason enough to oppose such policy.
Second, compliance is impossible. Few students genuinely understand what constitutes AI—from extractive to generative and all variants of the two—indeed, few professors have any such capacity. Nobody agrees precisely what constitutes ‘artificial intelligence,’ because much machine learning is implementation of predictive mathematics that has existed for generations. The ‘magic’ of contemporary AI is a combination of centuries-old mathematics, decades-old algorithms, and modern computational power. So, it is nonsensical to require ‘disclosure of AI use,’ as if that were a determined something. It is not. For example, contemporary spell checking is mathematically AI. Nor will commercial products be honest about what is under the hood—from 2026 advertising, one should be forgiven for thinking next to everything is AI and half of it is quantum too.
Third, compliance ranges from awkward to bizarre. I spend a lot of time with my soulmate, and so I can’t imagine how many times I bounce ideas off of her in various stages of development. It would be ridiculous for anyone to require that I make a note of each of those, logging my ‘spouse inquiries,’ and turn it in with any final product. (If god tries, I’ll tell her off too. What is between Hilary and me is between Hilary and me, thank you very much.) Similarly, it would be absurd for anyone to have to log every dictionary and encyclopedia consultation of a project’s lifespan. Indeed, it would be outrageous to require one to, ‘jot down every written source you consulted, however used, in producing this work.’ As I’ve pointed out elsewhere, everything I know was pretrained on the content of other humans. Should I write that down too? Generative AI is simply another—and, carefully used—a better resource than humans have ever had. Those who haven’t figured that out yet will… or they will become leftovers in tomorrow’s information economy. A requirement that one log every catalyst of a final product is absurd.
So, that makes for a perverse incentive (an ‘honesty tax’), definitional collapse, and epistemological absurdity—good enough reason for me.
Leave a Reply