iiinova — where your digital self keeps evolving

Jeremy Sheff @jnsheff.bsky.social · 15/09/2026

"A familiar future, yet stupid." About as good a description as any.

Reposted by Jeremy Sheff

Kate Klonick @klonick.bsky.social · 15/09/2026

People 2014-2024: “We don’t want robots reviewing our content! We demand human review!” People in 2026: “Holy shit, did you know humans are reviewing our content!? How creepy! We only want robots!!”

59721

Jeremy Sheff @jnsheff.bsky.social · 14/09/2026

I'll be presenting a keynote at Clemson's Law and Society Symposium next week. The symposium addresses legal and philosophical issues raised by AI. My lecture is organized around the question: "What Is Intelligence For?" If you're in the area, I hope you'll attend. jeremysheff.com/2026/09/14/a...

jeremysheff.com

AI Law And Society at Clemson – Jeremy Sheff

I'll be presenting a keynote lecture at Clemson's Law and Society Symposium On September 23, 2026. The symposium addresses legal and philosophical questions around AI, and my lecture is organized arou...

022

Reposted by Jeremy Sheff

Jess Miers 🦝 @jmiers230.bsky.social · 13/09/2026

I did a somewhat deep dive into the recent OpenAI and Anthropic "hacks" for an upcoming interview. IMO, the AI companies are better off running with the "rogue AI" story because the real story is actually pretty embarrassing. Like making your Wi-Fi password "wifi" embarrassing. 🧵

341836794

Reposted by Jeremy Sheff

Jess Miers 🦝 @jmiers230.bsky.social · 13/09/2026

And of course OpenAI and Anthropic are running with it too. Rogue AI narratives easily let them off the hook for their sloppy (and embarrassing) development decisions. While also making it look to their eager investors like they're inventing god...

250983

Reposted by Jeremy Sheff

Pwnallthethings @pwnallthethings.bsky.social · 12/09/2026

This is a good example of how you need quite a lot of detail of how these systems *actually* work to understand what the risks are, and how over-indexing on sci-fi can lead you a bit astray on what the future looks like

342120492

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

Adding, if you made it this far, you may be interested in "Artificial Intelligence Law," the new textbook @klonick.bsky.social and I will be releasing later this year. artificialintelligencelaw.org

artificialintelligencelaw.org

Artificial Intelligence Law

Technology, Law, and Policy

2325

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

This is a unique and precious moment in which the regulatory environment for AI is still up for grabs. We ought to approach the moment with a full appreciation of *all* our options, the advantages and disadvantages of each, and the vested interests of those advocating for any one option over others.

2337

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

Their safety proposal-costly private monitoring, top-down regulation with costly compliance that only the largest companies can manage-should all be read in light of those interests.

1303

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

Frontier AI companies have a *very* strong interest in obtaining the same kinds of immunities against tort liability that prior disruptive technologies have received, and in insulation against competition that might erode their current market-leading positions.

13511

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

No such privileges have yet been afforded to AI technologies. Cases currently in litigation are going to decide what their tort liability will look like, unless legislators provide them with statutory immunity first (as some seem interested in doing).

1234

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

The result has been that US tech companies have successfully internalized the benefits of disruptive technologies while externalizing the costs of the harms those technologies can cause.

13810

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

And so US tech industries have become accustomed to privileged insulation against tort liability for the harms their technologies cause (or enable users to cause): heightened knowledge requirements for IP infringement and negligence liability, the DMCA safe harbor, CDA Section 230.

1256

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

US technology law over the past half century has been wary of tort liability. We worry about stifling innovation, about foregoing great technological benefits to avoid less serious technological risks, and more recently about geopolitical competition.

1253

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

But command-and-control regulation is not the only regulatory tool available. To force frontier AI companies to internalize the externalities of social or systemic risk, the obvious policy lever is tort liabity: make AI companies liable for the harms their models cause (or enable users to cause).

25011

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

The classic solution to the collective action problems created by markets is regulation. And notwithstanding the danger of regulatory capture, such regulation seems inevitable for frontier AI models in the short to medium term.

1345

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

Even if a monitoring target would take a hit to their reputation for safety upon switching external monitors, the lesson we've learned again and again is that when put to the choice between safety and market success, frontier AI companies will not choose safety. Another collective action problem.

1306

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

So even though organizations like METR have internal ethical controls-funding restrictions, standards and practices, disclosures-if a frontier AI company threatens to take away their access (let alone all their free tokens), they will be put to the decision of what exactly their ethics are worth.

1294

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

There are no such licensure requirements for AI safety monitoring (yet).

1302

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

And even so, such breaches do happen when professionals in a competitive market look the other way to avoid losing a client, hoping that they simply won't get caught.

1302

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

This is why professions such as law and accounting have rules like the code of professional responsibility and generally accepted accounting principles, the breach of which can subject the professional to liability and even loss of license and livelihood.

1413

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

Knowing this, external private monitors who care about maintaining their access and monitory role will avoid looking too close or pushing too hard against risks that threaten the monitoring target's core objectives.

1403

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

The problem with private monitoring is a classic collective action problem: if any one private monitor identifies risks that the target does not want to abate, the target can simply contract for a more pliant monitor.

26817

Jeremy Sheff @jnsheff.bsky.social · 12/09/2026

The first element of this strategy (external private monitors against systemic social risk) was tried in the aughts with credit rating agencies in the mortgage-backed securities market. It failed. darioamodei.com/post/we-must...

darioamodei.com

Dario Amodei — We Must Pace the Frontier

111641

Jeremy Sheff @jnsheff.bsky.social · 25/08/2026

Explaining Markov models on the first day of Artificial Intelligence Law. With @klonick.bsky.social .

Classroom demonstration of Markov models.

2201

Jeremy Sheff @jnsheff.bsky.social · 20/08/2026

New Paper Alert: The Japan Trademarks Dataset: An open-access, comprehensive dataset on the trademark registration system of Japan. Paper forthcoming in the Journal of Law and Empirical Analysis (JLEA) jeremysheff.com/2026/08/20/n...

jeremysheff.com

New Paper Alert: The Japan Trademarks Dataset – Jeremy Sheff

The Japan Trademarks Dataset is now available on Zenodo.

033

Reposted by Jeremy Sheff

Larry Glickman @larryglickman.bsky.social · 02/07/2026

OTD in 1964, President Lyndon B. Johnson signed the Civil Rights Act. Here is a photo of LBJ handing a pen to Martin Luther King, Jr, at the signing ceremony. As the caption says, this was "the strongest civil rights legislation since Reconstruction Days."

Photo of LBJ handing a pen to MLK after signing the law.
HANDSHAKE FOR KING--President Johnson shakes hands with the Rev. Martin Luther King after handing him a pen during the signing of the bill, the strongest civil rights legislation since Reconstruction Days.

7408175

Reposted by Jeremy Sheff

Mayor Zohran Kwame Mamdani @mayor.nyc.gov · 11/06/2026

SPEECHLESS. LFGK

209196292185

Jeremy Sheff @jnsheff.bsky.social · 11/06/2026

Illegal fireworks launch from the sidewalks to burst over the rooflines of Harlem, orange and blue.The NYPD cruiser double-taps its siren at the screaming revelers spilling out into the streets, and its bullhorn deadpans, "Knicks in five." New York loves New York.

040

Reposted by Jeremy Sheff

Josh Marshall @joshtpm.bsky.social · 11/06/2026

It’s 2 against one, the spurs and the refs against the Knicks. Wtaf.

2113417

Jeremy Sheff @jnsheff.bsky.social · 06/06/2026

New Yorkers propose edits to AI company's subway ad campaign.

A print display ad for Genspark.ai where the letter "d" in the slogan "One prompt, job done" has been covered with a decal of the letter "g".

041

Reposted by Jeremy Sheff

Justin Levitt @justinlevitt.bsky.social · 03/06/2026

On June 8, 2023, the Supreme Court told Alabama to draw an equitable congressional map. Alabama flatly refused. Yesterday, SCOTUS validated Alabama's decision. "States can just ignore SCOTUS rulings hoping the Court will change its mind" is a pretty odd position for a court of law to endorse.

11310

Jeremy Sheff @jnsheff.bsky.social · 13/05/2026

Important thread. mRNA-based treatments are, as predicted, showing promise as a clinical tool to prevent or significantly delay cancer deaths. Addressing honest but underinformed skepticism while defending good science against bad-faith charlatans will be crucial to realizing that promise.

121

Reposted by Jeremy Sheff

Niko Bowie @nikobowie.bsky.social · 03/05/2026

The simplest long-run solution to the gerrymandering wars is a federal law expanding the House and electing members by proportional representation. Alabama has more Democratic voters than Hawaii. Massachusetts has more Trump voters than Mississippi. None of these voters deserve to go unrepresented.

313756

Jeremy Sheff @jnsheff.bsky.social · 01/05/2026

Chris brings up an important piece of SCOTUS reform that I’ve written about before. The judges who grant cert should *not* be the ones who decide the merits. Ideally the cert bench would choose cases for the *next* merits bench *before* its members are selected. cardozolawreview.com/wp-content/u...

cardozolawreview.com

0101

Reposted by Jeremy Sheff

Sarah G. @sglassmeyer.bsky.social · 01/05/2026

attn law profs/academics …saw this on LinkedIn - Clearinghouse of calls for papers docket.richardsonresearch.org made by U Hawaii

docket.richardsonresearch.org

The Docket - Law School Calls for Papers

Find calls for papers from law journals, conferences, and workshops.

012

Reposted by Jeremy Sheff

Adam Serwer @adamserwer.bsky.social · 29/04/2026

The "reactionary colorblindess" of the right-wing justices has reached its logical conclusion in finding that it is racist NOT to let Louisiana—and any other state for that matter—discriminate on the basis of race in voting. (gift link) www.theatlantic.com/ideas/2026/0...

What we can expect in the aftermath of this ruling is for more Republican-controlled states to implement discriminatory maps and call them partisan so they can pass legal muster. In practical terms, this will likely mean fewer nonwhite representatives in Congress. Diminishing the power of minority voters may also allow the Republican Party to continue on its path from reactionary color-blindness to more overt racism, safe in the assumption that it will not have to answer to constituents who oppose such racism because they are its targets. There is little risk in attacking people who lack the power to remove you from office.

6235941168

Jeremy Sheff @jnsheff.bsky.social · 29/04/2026

John Roberts has been happy to serve as the respectable, mild-mannered face of revanchist American racism for my entire life. I'm sure he imagines himself the clear-eyed, even-handed steward of a racially fractured society. But I doubt he can imagine a society where that isn't his prerogative.

67437

Jeremy Sheff @jnsheff.bsky.social · 25/04/2026

A famous CompSci conjecture involves problems that are difficult to for computers to solve but whose solutions are easy for computers to verify. Workplace AI seems to have created the inverse: problems that humans can easily solve but whose computer-generated solutions are hard for humans to verify.

en.wikipedia.org

P versus NP problem - Wikipedia

074

Jeremy Sheff @jnsheff.bsky.social · 22/04/2026

And now there are AI hallucinations in a bankruptcy court filing by lawyers at S&C. These lawyers' clients have to wonder what exactly it is they're paying for. www.nytimes.com/2026/04/21/n...

nytimes.com

A.I. ‘Hallucinations’ Created Errors in Court Filing, Top Law Firm Says

000

Reposted by Jeremy Sheff

jamelle @jamellebouie.net · 12/04/2026

more seriously congrats to the hungarian people for kicking that asshole out

142227492560

Reposted by Jeremy Sheff

Alexander Clarkson @aphclarkson.bsky.social · 11/04/2026

This won't kill this technology, it just makes pharma a sector dominated by European, Chinese and Indian corporations

28991261

Reposted by Jeremy Sheff

Jonathan Ladd @jonathanmladd.com · 09/04/2026

Happy Surrender and Freedom Day! On April 9, 1865, Robert E. Lee and the Army of Northern Virginia surrendered to Ulysses S. Grant and the Army of the Potomac. home.nps.gov/apco/planyou...

6430571079