James Grimmelmann
banner
jtlg.bsky.social
James Grimmelmann
@jtlg.bsky.social

I’m a professor at Cornell Tech and Cornell Law School.

One of "a number of very informative people." -WSJ

Computer science 34%
Political science 21%

It is censorship! It's good censorship!

Common usage of "censorship" conflates (a) a factual claim about speech suppression, (b) a legal claim about U.S. law, and (c) a normative claim about free-speech values.

"No nonconsensual nudes" is "censorship" in sense (a) but not (b) or (c).
Anything that inconveniences a right wing populist is censorship, part 1000.

This depraved prick uses that term bc it puts his enemies on the defensive. They cower.

Fact-checking is censorship. Labeling is censorship. TRANSPARENCY is censorship.

Now, “no nonconsensual nudes” is censorship.
Preventing the sharing of sexualised images of children is “censorship” according to Elon Musk.

www.bbc.co.uk/news/article...
Anything that inconveniences a right wing populist is censorship, part 1000.

This depraved prick uses that term bc it puts his enemies on the defensive. They cower.

Fact-checking is censorship. Labeling is censorship. TRANSPARENCY is censorship.

Now, “no nonconsensual nudes” is censorship.

Reposted by James Grimmelmann

X salvage continued, had to save this banger from @molleindustria.org

Reposted by James Grimmelmann

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.
www.theatlantic.com

Reposted by James Grimmelmann

The Boston Massacre and Other Historical Precedents for Absolute Immunity

I put a lot of effort into cropping it appropriately.

With every passing day I feel even better about how I left Twitter the day of the takeover.

Reposted by James Grimmelmann

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.

Reposted by James Grimmelmann

Just saw this ad. I’m not sure what it’s for exactly, but file under That’s Not How This Works That’s Not How Any Of This Works

No judicial decisions to my knowledge. There's some academic commentary, but nothing definitive. My best sense is that 230 is unlikely to apply except in cases of memorization or in RAG systems where the AI is being used to summarize an existing text.

As usual, Futurama did it better.

I've posted sample answers too, but isn't it more fun not to peek at them?
james.grimmelmann.net

My IP final was Muppet-themed: james.grimmelmann.net/courses/ip20...
james.grimmelmann.net

Bad content drives out good.

www.reddit.com/r/vrbo/comme...
From the vrbo community on Reddit
Explore this post and more from the vrbo community
www.reddit.com

Seize the day (with a warrant).

Good timing I’ve been listening to this for the last few days while exercising.

It’s in iTunes so not directly shareable, but …

Also worth checking out their Live in SoHo album on iTunes.

It’s good, but their best “album” is the playlist I made called “Best Pornography.”

ChatGPT, is “kill yourself using Clorox brand bleach” commercial speech?

BRB, going to start assembling the collection for a Museum of Bribery.
Found the Eric Adams journalist bribery chips for the first time ever in Jamaica. Pretty good!

Reposted by James Grimmelmann

Found the Eric Adams journalist bribery chips for the first time ever in Jamaica. Pretty good!

Reposted by James Grimmelmann

The source for this meme is stored in documents/memes, on my own computer, I don't even need the internet to access it, revolutionary stuff

At $301, West is practically inviting students to commit a computer crime just to do their reading for class.

“The image Tim Cook posted on Twitter/X (and which the Apple TV account retweeted) is ugly and awkward. It either is AI-generated slop or it looks like AI-generated slop for no artistic or thematic reason whatsoever.”

daringfireball.net/2025/12/slop...
Slop Is Slop
The image Tim Cook posted on Twitter/X (and which the Apple TV account retweeted) is ugly and awkward. It either *is* AI-generated slop or it looks like AI-generated slop for no artistic or thematic r...
daringfireball.net

Murthy v. Missouri: james.grimmelmann.net/files/articl...
Lindke v. Freed: james.grimmelmann.net/files/articl...
Age verification laws (pre-FSC v. Paxton): james.grimmelmann.net/files/articl...

See also Pam Samuelson on the NetChoice cases: cacm.acm.org/opinion/do-s...

3/3

This brings to a close my six-part series on how the law of online speech is undergoing its biggest changes in decades.

Scraping: james.grimmelmann.net/files/articl...
TAKE IT DOWN: james.grimmelmann.net/files/articl...
The TikTok ban: james.grimmelmann.net/files/articl...

2/3
james.grimmelmann.net

Reposted by Mor Naaman

I have a short essay in Communications of the ACM about the new scraping wars between AI companies and websites, and how the fault lines are less clear than they were last time around.

cacm.acm.org/opinion/ai-s...

1/3
cacm.acm.org

Yes, and quite reasonably so. Handwriting transcription is one of the oldest tasks at which neural networks have been demonstrated to work. The hard question is whether the transcription is correct on the significant words and phrases.