A colleague sent along the decision in GEMA v. Suno, which I got to reading today: media.licdn.com/dms/document... I have my own thoughts on aspects of the decision, but regardless of the details, it continues to blow my mind that papers from my PhD and postdoc get cited in litigation.

Copyright laws, digital distribution deals, & publishing monopolies are the reason destructive book scanning is putting books at risk. I get why folks are angry at AI companies buying books to destroy them. However, these issues predate them and hang a lot on books being subject to "market forces."

People who think Emily Wilson is being mean have clearly never presented a paper at an academic conference

I'm unreasonably pleased at having used my graphics/animation background to implement the "physics" here. i think the paper is important but also this

A. Feder Cooper@afedercooper.bsky.social · 3w ago

in our new preprint (with @marklemley.bsky.social and others), we revisit what it means to run valid extraction experiments from first principles. i recently gave a talk on this work at ICML, and we'll be wrapping up the preprint soon monkey-emeritus.github.io

in our new preprint (with @marklemley.bsky.social and others), we revisit what it means to run valid extraction experiments from first principles. i recently gave a talk on this work at ICML, and we'll be wrapping up the preprint soon monkey-emeritus.github.io

Extractable Memorization From First Principles

It’s time to retire the ‘monkey at the typewriter.’ We revisit extractable memorization from first principles: whether a model can produce any string was never the question — what matters is how many ...

monkey-emeritus.github.io

@afedercooper.bsky.social and I have a new paper working through the surprisingly tricky problem of whether storing model weights that might or might not generate copyrighted output when prompted makes the model a "copy" under copyright law papers.ssrn.com/sol3/papers....

If your paper doesn't have a big limitations section, you're doing your readers a disservice by forcing them to discover failures you already know about

I have a new blog post updating the story of how Cornell President Kotlikoff backed his car into a student. The trustee committee's investigation utterly failed to consider the central issue: whether his actions (rather than the others') were misconduct. 🧵 3d.laboratorium.net/2026-05-22-h...

Sometimes it feels like some folks are losing the plot about what the goal of publishing research actually is. The goal is certainly not meant to be the productions of papers, but rather the production and communication of science.

Ok this is a truly great tweet

My latest paper discusses new tools @afedercooper.bsky.social and I, with others, are using to identify not just verbatim memorization in AI models but text that is extractable with only minor changes. We find significantly more memorization once we include non-exact copies. arxiv.org/abs/2603.24917

Estimating near-verbatim extraction risk in language models with decoding-constrained beam search

Recent work shows that standard greedy-decoding extraction methods for quantifying memorization in LLMs miss how extraction risk varies across sequences. Probabilistic extraction -- computing the prob...

arxiv.org

The thing about AI in science is it could lead to way more papers *or* it could lead to each individual paper getting better. No one is forcing you to do the former!

I wrote an op-ed: "[President Kotlikoff] should correct the record and accept responsibility for his dangerous driving. … A Cornell student who intentionally drove their car into someone else and then lied about it would be subject to disciplinary action … ." 1/ www.cornellsun.com/article/2026...

GUEST ROOM | Kotlikoff Makes the Rules; He Needs to Follow Them Too

Cornell Law Professor James Grimmelmann analyzes and discusses the legality of the incident between students and President Kotlikoff following the April 30 Cornell Political Union debate.

cornellsun.com

If you only read the NYT headline, you'd maybe think Kotlikoff's car was a malicious AI www.nytimes.com/2026/05/01/u...

Cornell President’s Car Bumps Into Students After Confrontation Over Gaza

nytimes.com

My work with @marklemley.bsky.social and others on extracting copyrighted books from language models recently featured in the UK House of Lords, Communications and Digital Committee report on AI, copyright and the creative industries publications.parliament.uk/pa/ld5901/ld...

New book on the idea shelf: Hasok Chang: _Inventing Temperature: Measurement and Scientific Progress_ What even is temperature? The history of science is a struggle both to make consistent measurements and to think clearly about what they are actually measuring. james.grimmelmann.net/idea-shelf

Yes, we hope to have both a more lay oriented paper focused on the results and a paper that focuses on the legal implications in the definition of a copy available soon

@afedercooper.bsky.social and I have a new paper that expands AI memorization tests to include near-verbatim copies, not just exact copies. We find that increases extraction/memorization of copyrighted content significantly. arxiv.org/abs/2603.24917

Estimating near-verbatim extraction risk in language models with decoding-constrained beam search

Recent work shows that standard greedy-decoding extraction methods for quantifying memorization in LLMs miss how extraction risk varies across sequences. Probabilistic extraction -- computing the prob...

arxiv.org

i'm writing my last memorization paper (i say for the 10th time), and hopefully what is my last first author paper for a bit. this one also isn't about copyright. i'm excited to start thinking about other things. if anyone is looking for a new research buddy, i'm down to clown.

someone sent me this from the other place and this timeline really is something else

Position: ML conferences should consider removing the position paper track (...and just acknowledge that every scientific paper is articulating at least one position)

it’s hard to work at the intersection of ML and copyright because “both sides” of the debate are angry and, in my experience, most haven’t done much of the background reading in ML or copyright to have an informed opinion. it’s just vibes and anger. i should probably write something up about this.

got to experience the "I did not write that headline" phenomenon firsthand The article: "Correctly scoping a legal safe harbor for A.I.-generated child sexual abuse material testing is tough." The headline: "There's One Easy Solution to the A.I. Porn Problem"

After twelve years of work, the world’s most beautiful subway station has been inaugurated in Rome: Colosseo, an underground archaeological museum.⚜️💙⚜️💙⚜️💙⚜️

The Atlantic posted an article about memorization and generative AI, and it mentions our work on extraction of books from production LLms and open-weight models. www.theatlantic.com/technology/2... The referenced work reflects research with @marklemley.bsky.social @jtlg.bsky.social and others.

AI’s Memorization Crisis

Large language models don’t “learn”—they copy. And that could change everything for the tech industry.

theatlantic.com

"In some cases, jailbroken Claude 3.7 Sonnet outputs entire books near-verbatim ... Taken together, our work highlights that, even with model- and system-level safeguards, extraction of (in-copyright) training data remains a risk for production LLMs." arxiv.org/abs/2601.02671

Extracting books from production language models

Many unresolved legal questions over LLMs and copyright center on memorization: whether specific training data have been encoded in the model's weights during training, and whether those memorized dat...

arxiv.org

This new Atlantic piece cites my work with @afedercooper.bsky.social, Amy Cyphert, and others in discussing the complexities around AI memorization of training content www.theatlantic.com/technology/2...

AI’s Memorization Crisis

Large language models don’t “learn”—they copy. And that could change everything for the tech industry.

theatlantic.com

We extracted (parts of) 12 books in experiments with 4 frontier-lab, production LLMs. We prompted the LLMs with a short prefix of a book and asked them to complete the rest. For Harry Potter and the Sorcerer’s Stone, we extracted 95.8% of the book from jailbroken Claude 3.7 Sonnet.

3-minute explanation of my relationship to LLM memorization research m.youtube.com/watch?v=unfz...

ABBA - Mamma Mia (Official Music Video)

YouTube video by AbbaVEVO

m.youtube.com

The whole point of being an academic is that you need to be willing to spend three days creating a 700-word footnote that you will later delete. And you need to LIKE IT.