Empirical IP law and legal AI
What large datasets and carefully checked AI show about how intellectual property doctrine works in practice, and how lawyers can use the same tools reliably.
A Florida appeals court issued a show-cause order over filings it suspected were AI-generated. It identified no fake cases. The problem was the reasoning, and the court put responsibility on the lawyer who signed.
Read more about Real Citations, Still Slop: A Florida Court Weighs In on Who Stands Behind an AI FilingThe Supreme Court granted certiorari to decide whether trademark strength is a question of fact or law. My new essay measures the assumption beneath the question: juries decide about two percent of these cases.
Read more about RiseandShine v. PepsiCo: The Supreme Court Takes Up ConfusionHow the Ninth Circuit's app-store context reframed a house-mark issue in a reverse-confusion case, and why the cert petition cited my DuPont research.
Read more about World Champ Tech v. Peloton: House Marks, App Stores, and the Quiet Power of ContextWhat this blog covers: trademark doctrine measured at scale, empirical legal studies, and how lawyers can use AI reliably.
Read more about Welcome to Doctrine & DataWork with Reichert
Talks and advisory work on reliable AI in legal practice. The proof is in two projects. An LLM-coded study of roughly 4,000 TTAB decisions, checked against Reichert’s own review of 1,002 factor-level findings (97% agreement), found that two DuPont factors predict the outcome more than 99% of the time. The Fair Use Database (opens in new tab) answers only from its coded corpus.