Danny Wilf-Townsend

To my knowledge, 2 papers of mine have been put through LLMs and reposted on arXiv/ResearchGate as novel work. 1 was published in a Nature subsidiary (+ verbatim copies figures from my 2021 paper). The 2nd popped up today in my arXiv digest, is a sloppified version of a paper I put up in early July.

I have mostly switched to lurking here but some words are so good they must be shared. “Surface the viscerality of parking and of religion”

After reading this, I wanted to know what the numbers would look like with more recent draft law review articles posted to SSRN. So I put a sample of 200 of them through Pangram 4, and found 57 papers with some AI use, 36 of which had >5% of their text from AI: wilftownsend.net/how-many-rec...

Excited to be back in New Haven for the academic year, visiting at YLS. Let me know if you come to town! So far I've had pizza three times in ten days, it's going to be work to keep that pace up.

Very happy to see this study finally made available publicly. In my many conversations with law professors about AI over the last few months, this study is the one that I have found most consistently helps people update their worldview about AI capabilities.

This is tangential, but I wonder how the idea / expression distinction will hold up over time with generated AI. Expressing an idea used to be costly, typically taking effort and often some creativity. What happens when ideas serve as prompts and expression is cheap or free, and easily iterated?

Happy to see a cameo here from one of my favorite tests in all of the law: whether a procedural rule is really a procedural rule depends on whether it "really regulates procedure."

some good climate/energy news: * 96% of new US power capacity was carbon-free in 2024 (56 gigawatts!) * 2025 included the first month ever when 51% of power on the U.S. grid was carbon-free * The golbal trend is overwhelming: The world is now investing more $ in clean energy than fossil fuels

I have a new post out in @lawfaremedia.org today about continual learning, the goal of many AI developers to build tools that can learn from their users. That technology could have many uses, but also will challenge existing ways we are trying to regulate AI. www.lawfaremedia.org/article/when...

Regulating static AI models is already difficult and if AI tools will become ones that can learn, regulations will need to adapt quickly. @dannywt.bsky.social explores what new regulatory approaches could look like in a future where change is common and comes fast.

A thoughtful thread on the Netflix / Warner Bros merger. I think the points about consumer preferences are particularly important — it’s sometimes hard, but often important, to tease apart when law and policy arguments are inflected by different preferences about product features

An update for Sonnet 4.5, released last week: it scored 60.2% on my final exam (with extended thinking on, 54.4% without it). That's a big step up (~20 percentage points) from Opus 4.1's scores, and puts Sonnet 4.5 close to, if slightly behind, other lead models. On a human curve, that's ~ an A-/B+

One other note: across the five exam answers and dozens of answer evaluations generated here, I did not notice a single hallucination. This test wasn't designed to measure hallucination rates, but it's consistent with the general sense that they have dropped significantly

Our office is again hiring one or more attorneys for a one-year fellowship to work directly with the Illinois Solicitor General and her team, beginning in August/September 2026. www.governmentjobs.com/careers/ilag...

A nice quick read from my colleague @JonahPerlin about an issue that I see a lot of people oversimplifying: whether an attorney's use of a generative AI tool waives privilege. This is an area where I'm very interested to see how the law develops. news.bloomberglaw.com/us-law-week/...

In a stunning moment of self-delusion, the Wall Street Journal headline writers admitted that they don't know how LLM chatbots work.

Judge Alsup has the first true opinion on fair use for generative AI in Bartz v. Anthropic. He holds that AI training is fair use, and so is buying books to scan them, but that downloading pirated copies of books for an internal training-data database is not fair use. 🧵

I think this is one of the more common mistakes I see with people trying AI—the idea that if you go to a free chatbot, quickly run a question by it, and it does a bad job, then you've learned that AI cannot do a good job on that question.

Significant ruling in one of the big algorithmic price-fixing lawsuits going on right now: www.reuters.com/legal/govern...

A good thread on a big new generative AI / IP lawsuit—Disney and Universal vs. Midjourney

ChatGPT is down but The Museum of English Rural Life still stands, proving once again that Silicon Valley cannot compete with the history of rural England and its people.

I've had a few recent conversations with judges and law profs who haven't tried generative AI, or have only used it for a few minutes to write a poem or other trivial fun. After a few people asked me about how to start, I thought I'd write up my suggestions: www.wilftownsend.net/p/some-ideas...

After years of studying AI & law here’s my rule of thumb: Lawyers should only use AI only when they can confidently assess, adapt & explain its output without engaging in deep, independent thinking about the core legal/factual issues. More in my new essay: papers.ssrn.com/sol3/papers....