Daniel W. Linna Jr.

Law must be a part of computer science curricula Agreed, @mireillemoret! Excellent @CACMmag piece! So much important CS+Law work to be done! We're onboard at @NorthwesternU w interdisciplinary CS+Law courses in both @NorthwesternEng & @NorthwesternLaw! cacm.acm.org/magazines/2021…

Website and registration link for AI for Legal Reasoning and Adjudication at @NLawBizTech @NorthwesternLaw: sites.northwestern.edu/lawandtechnolo…

How are judges and lawyers using AI in courts and law practice? How capable is AI for legal reasoning and adjudication? What are the possible benefits and risks of using AI for legal tasks? How might AI transform courts, dispute resolution, access to the law and legal services, and the rule of law? An outstanding lineup of judges, practitioners, and academics will speak about these topics and more on Thursday, April 2, 2026 (8:30 am to 5:00 pm) in Thorne Auditorium at Northwestern Pritzker School of Law. Register to join us in person or on Zoom.

People joking that NYC are going to ban free legal advice from books next... Don't realize NYC already banned free legal advice from newspapers and radio in 1936! via our newest substack: newsletter.pessimistsarchive.org/p/nyc-once-ban…

Claude's take on OpenAI being sued for unlicensed practice of law: "The real threat to the insurance litigation model isn't that AI gives bad legal advice. It's that it gives adequate legal advice at zero marginal cost to people who were previously priced out of fighting back. That's what $10.3 million in damages is really trying to price in."

This winter I co-taught a grad seminar on when & how LLMs can stand in for human subjects in social science research, to a mix of CS & social science students. It was fun (albeit w/occasional friction) to talk methods w/such a diverse group. Reading list: statmodeling.stat.columbia.edu/2026/03/10/new…

Join us on next Friday March 6 for a timely conference on Artificial Intelligence and Law at CU Law School's @SiliconFlatiron center. Leading AI experts will discuss ways to nudge AI use, and data centers, down socially beneficial paths. siliconflatirons.org/events/annual-…

Join me for an Artificial Intelligence Conference at @SiliconFlatiron @ColoLaw on Friday March 6! "Using AI Well — and Building a Data Infrastructure to Support It" This will be a fascinating and timely discussion with leading AI experts. Register here: siliconflatirons.org/events/annual-…

🚨 LAST CALL: Free Legal Ethics CLE Alert 🚨 From the War-Room to the Bench: Navigating AI in the New Legal Frontier linkedin.com/posts/mcpalmer… REGISTRATION: beacon360.content.online/xbcs/S2692/cat… #LegalTech #AI #LawFirms #CLE #LegalEthics

On this week Pioneers and Pathfinders episode Stephen Poor welcomes returning guest, Dan Linna, director of law and technology initiatives and senior lecturer at Northwestern Pritzker School of Law and the McCormick School of Engineering. Listen here: ow.ly/VBGT50Y8Yp9

Great @dhadfieldmenell thread about software projects, which is relevant for AI for Law. In law, in addition to the difficulties of specifying everything up front, we lack metrics for quality and evals for outputs. Law has not had a quality movement. Law is not evidence based. AI is forcing change. People-Process-Data-Tech approaches are key. Designing AI for Law interfaces to require user engagement is crucial. These domain-specific pieces are missing in general purpose AI tools. For the foreseeable future, domain-specific approaches are needed to increase accuracy to reasonable levels.

Excited to share my latest paper, co-authored with 3 computer scientists, about how AI e-discovery software can either hide or help to expose Brady evidence, depending on configuration. papers.ssrn.com/sol3/papers.cf…

If the wildest AI bulls are right and all knowledge work is about to disappear fast, that implies a lot for today: nobody should be using AI to learn job skills, or automate their daily workflows, or anything else, since the work world is about to become unrecognizable 1/

Lawyers and law professors should be preparing for the impact of AI on the profession and rule of law. At the same time, we should be skeptical about the claims of the leaders of AI labs. This matters because where we predict AI will be in 2, 5, 10 years and more should influence our learning objectives for students today. Our predictions about AI also heavily influence policy, law, and scholarship. We should prepare for the (small?) possibility of super intelligent AI in the near term, but many (most?) sober computer scientists argue that it is much further off. For example, I find Andrew Ng's predictions to be based in evidence rather than appeals to exponentialism. Gary Marcus, while too skeptical, identifies important evidence about the limitations of LLMs for discussion. Yann LeCun has also concluded that LLMs are a dead end to artificial general intelligence. Nonetheless, existing AI tools, while limited, are quite capable for augmenting many legal tasks, and automating some, especially when the tools are engineered into a system designed for those specific legal tasks. Andrew Ng: nbcnews.com/tech/innovatio… Gary Marcus: garymarcus.substack.com/p/the-last-few… Yann LeCun: fastcompany.com/91462273/yann-…

Deal on @Instacart for @FreshThymeFM $10 off on $50. Put $50 in cart. No discount. Go back and click on link: "Here is a coupon for $10 off a restaurant order!" Misleading. Deceptive. Unfair.

One of the most common mistakes people make when evaluating the pace of AI research is to look at progress on one type of task and extrapolate it to all tasks that humans can do. AI progress is extremely vertical-specific. In the past year, verifiable domains and in particular code have shown fast progress, which does not extend to other domains. This is because the main driver of AI capabilities remains, to this day, the memorization and operationalization of past data, which can be generated in unlimited quantity in the case of verifiable domains.

Today we are releasing OpenExempt, a new framework and benchmark for diagnostic evaluation of legal reasoning in language models. OpenExempt rethinks evaluation by shifting control to the benchmark user. Moving beyond static datasets, the OpenExempt Framework generates complex legal reasoning tasks on demand, where each scenario is dynamically shaped by a user-defined configuration to target their evaluation goals. OpenExempt computes gold solutions for each task using expert-crafted symbolic representations of relevant U.S. federal and state statutes, an approach inspired by legal DSLs such as Catala and Accord. Using this framework, we construct the OpenExempt Benchmark, a diagnostic benchmark with 9,765 samples across nine evaluation suites, designed to carefully probe model capabilities through controlled task variation. OpenExempt was developed by an interdisciplinary team with my coauthors: Sarah Lawsky (@sarahlawsky), Rajiv Jain, Daniel W. Linna Jr. (@DanLinna), and Kristian Hammond (@KJ_Hammond). We release OpenExempt to the public under a permissive license (CC BY 4.0) to support further research and encourage collaboration between the legal and NLP communities. • Evaluation: If you are evaluating language model reasoning internally, easily incorporate the OpenExempt Benchmark to complement these efforts. • Customization: If the standard benchmark doesn’t fit your evaluation goals, the OpenExempt Framework lets you construct a custom benchmark that does. • Collaboration: We are interested in research collaborations focused on substantial extensions to this work. If that’s you, reach out and let’s discuss. Read the paper: arxiv.org/abs/2601.13183 Run the code: github.com/servantez/Open… Access the benchmark: huggingface.co/datasets/Sergi… This work was supported by the Center for Advancing Safety of Machine Intelligence (CASMI) at Northwestern University.

To Michigan ,all of our former Michigan Men players , Warde Manuel our Regents, President and our fans. Thank you for the honor of serving. I will always call Michigan my football home. God Bless you with true meaning in your lives of service to others, a deep faith in God and the love of your families. Please positively support Kyle and the players. Be true GoBlue.

The Guild Strikes Back? Lawyer protectionism is a terrible look. It will be less and less effective in the future. UPL is crumbling. New regulatory structures that are faithful to competition law and theory are needed. There will be plenty of legal tasks to be done in the future, and plenty of opportunities for lawyers and other professionals to contribute to improving access to law, legal services, and justice and the rule of law. The economics and who benefits will change. Not all lawyers will benefit and it will not be only lawyers. Embracing AI and other emerging technologies is the only way to remain relevant and serve society.

Notre Dame football has Community Notes working overtime.