UC Berkeley lost again. On March 26, 2026, the U.S. Patent Trial and Appeal Board ruled that the Broad Institute, not Berkeley, invented CRISPR Cas9 gene editing in human cells first. It was the latest defeat in a fight that has run for more than a decade and left two Nobel laureates caught in the middle of a dispute they did not personally litigate.
The case is worth understanding even if you have never touched a pipette. At its center is a question every researcher eventually runs into when working with industry, a lab partner, or even a rival team down the hall: who owns an idea, and at what point does a hunch become an invention?
The Timeline
In June 2012, Jennifer Doudna at UC Berkeley and Emmanuelle Charpentier published the first paper describing CRISPR Cas9 as a programmable gene editing tool. The experiments worked in test tubes, and the paper went on to become one of the most cited in scientific history, according to Reuters.
That December, Berkeley filed a provisional patent application. It described the core components of CRISPR Cas9, but it had not yet shown the system working inside human cells, IPWatchdog reports. That same month, Feng Zhang at the Broad Institute filed his own application. Zhang paid for expedited review and, unlike Berkeley, submitted detailed evidence that CRISPR worked inside human cells, according to Morrison Foerster.
In April 2014, the Patent Office declared an interference proceeding, a formal process meant to determine who invented CRISPR Cas9 for eukaryotic cells first, per Reuters. By February 2017, the PTAB had ruled in favor of the Broad Institute, finding that Berkeley’s 2012 application did not adequately describe how to make CRISPR work in human cells, according to IPWatchdog. The Federal Circuit affirmed most of that ruling in September 2018, and Berkeley appealed again, Morrison Foerster notes.
In 2020, Doudna and Charpentier won the Nobel Prize in Chemistry. The patent war continued anyway.
The case took a turn in May 2025, when the Federal Circuit vacated the PTAB’s decision and sent it back down. The court said the board had applied the wrong legal standard for conception. The real question was never whether Berkeley’s scientists were certain their invention would work; it was whether they had formed a definite idea that only routine skill remained to execute, IPWatchdog explains.
It did not matter. On remand this March, the PTAB reaffirmed its original ruling. Even under the corrected standard, the board found, Berkeley had not met its burden, according to Morrison Foerster. The scientists’ own lab notebooks expressed doubt about whether their approach would work in human cells, and the labs that eventually replicated the work successfully had capabilities beyond what an ordinary researcher possessed at the time, per IPWatchdog.
What the Case Actually Reveals
The 2026 ruling came down to one finding: Berkeley’s scientists did not have a definite, operable idea before the Broad Institute reduced the invention to practice. Their own records showed they were still figuring out the variables. For any researcher in a partnership with industry, that distinction has teeth. Who conceived what, and when? What proves it?
From 1996 to 2020, university research produced more than 495,000 inventions and 126,000 U.S. patents, and academic tech transfer has contributed an estimated $1 trillion to U.S. GDP over that span, per a GAO report released this February. Berkeley’s CRISPR patents are one contested piece of that much larger system of university research turning into commercial value.
The contracts that govern these partnerships vary more than most researchers realize. A 2025 study in Research Policy examined 484 university industry research contracts and found IP ownership, publication rights, and confidentiality terms differed sharply from one collaboration to the next. The study identified three common models: proprietary partnerships, controlled access agreements, and open science collaborations, with controlled access being the most common arrangement. One pattern held across the sample: the more public funding a project relied on, the less ownership and use rights industry partners tended to retain.
Four Lessons From Berkeley’s Loss
Each of the four gaps below appears somewhere in the record of this case: in the notebooks, in the filing dates, and in the broader question of how research partnerships and technology transfer offices are meant to prevent disputes like this one.
Lesson 1. Document conception rigorously. The gap between a hypothesis and an invention is the gap between winning and losing a patent. Lab notebooks need to record not just what worked, but the moment an idea became definite enough that someone else could execute it without further insight. Berkeley’s own notebooks expressed doubt about whether their approach would work in human cells, while the Broad Institute submitted evidence that its separate work already did.
Lesson 2. File before you publish, not after. Berkeley published first and filed second. The Broad Institute filed early and paid to have its application reviewed on an expedited track. That sequence, publish then file versus file then publish, is part of the record the PTAB weighed.
Lesson 3. Pre negotiate IP terms before the research starts, not after it succeeds. Background IP, what each party owns walking in, and foreground IP, what the collaboration creates, need to be defined in writing before anyone touches a bench. Canada’s NSERC guidelines make this distinction explicit and recommend settling it before collaborative work begins at all. Berkeley and the Broad Institute were not partners in a shared project, so this particular lesson applies less to their dispute directly and more to any researcher entering a joint arrangement where that clarity does not yet exist.
Lesson 4. Bring in your Technology Transfer Office early, not once a dispute is underway. TTOs exist to manage filings, draft sponsored research agreements, and keep everyone compliant with Bayh Dole. The University of Minnesota learned this the expensive way: a seven year storage IP dispute with Broadcom ended in an $85 million settlement in 2025, after which Minnesota moved to require quarterly IP audits on every major research project, according to AI Trademark Review.
Related Reading
- Conception vs. Reduction to Practice: The Legal Line That Decides Patent Priority – Two teams, one discovery, different outcomes. How U.S. patent law defines the exact moment an idea becomes an invention.
- Bayh Dole at 45: What Federally Funded Researchers Still Get Wrong About Ownership – Federal funding comes with rules most labs never read closely, until a dispute forces them to.
- Background IP vs. Foreground IP: The Contract Clause Most Research Partnerships Skip – One paragraph, defined early, can prevent years of litigation over who owns what a collaboration produces.
- When Industry Partners Restrict Publication Rights: The Microsoft OpenAI AGI Clause – What happens when a funding agreement gives one partner the power to decide what the other can say.
