@j-p-a.bsky.social with his latest installment of work on AI in rulemaking 👇

LLMs can, in a single exchange, produce incredibly sophisticated and utterly baffling responses. I have a new essay in the Harvard Journal on Legislation's Markup blog working through how agencies might responsibly use LLMs in the rulemaking process. journals.law.harvard.edu/jol/2026/04/...

"The common sense of mankind demands that law shall not stop with the punishment of petty crimes... It must also reach men who possess themselves of great power and make deliberate and concerted use of it to set in motion evils which leave no home in the world untouched." —Robert Jackson, Nuremberg

That’s wonderful. Connelly has “it.”

Bridget offers a useful corrective to the buzzwordification of how we talk about government—and, if we're not careful, how we actually govern. "We can likely make great progress in regulatory policy by letting algorithms into our loop, not the other way around."

Joe Tomlinson & @brendanmcgurk.bsky.social give an overview of how the UK is handling the intersection of AI and administrative law. www.yalejreg.com/nc/artificia...

Tara Aida notes that APA § 553 has a background assumption that human intelligence runs administrative processes. "If agencies over-rely on AI to carry out these procedural tasks, they threaten to undermine § 553’s goal of improving the quality of final rules." www.yalejreg.com/nc/ticking-t...

@jtillipman.bsky.social: "complex challenges arise when [AI tools] extend into discretionary functions, including core evaluative tasks, that federal procurement doctrine presumes a human decision-maker will perform." www.yalejreg.com/nc/abdicated...

Reeve Bull gets us up to speed on how Virginia has been using AI tools to engage in regulatory reform. Click here for an insider's view. www.yalejreg.com/nc/ai-empowe...

Elliot E.C. Ping works through whether chain-of-thought prompt engineering can solve the black box problem. www.yalejreg.com/nc/iterative...

@adamgcrews.bsky.social argues that courts can adapt to agency use of AI by looking to agency practice, rather than through judicial creativity and policy balancing. www.yalejreg.com/nc/agencies-...

Gilbert Orbea and Emily Froude of @democracyforward.org lay out three preliminary factors to guide courts when reviewing the use of AI in the regulatory process. www.yalejreg.com/nc/determini...

Jack Jones and Burçin Ünel of @policyintegrity.bsky.social offer that, whether an agency uses AI or not, if a "rule ignores key evidence, fails to address major concerns or alternatives, or offers inconsistent reasoning, it will be struck down as arbitrary." www.yalejreg.com/nc/do-large-...

Next, @j-p-a.bsky.social and John Lewis draw on recent reporting to argue that if "agencies setting machines loose to generate and justify regulatory proposals once seemed far-fetched, it no longer does" & that admin law standards can keep agencies accountable. www.yalejreg.com/nc/toward-mi...

Today in our AI & the APA symposium: @cary-coglianese.bsky.social discusses the place of those very confident-sounding chatbots in administrative decisionmaking. www.yalejreg.com/nc/ai-taxi-d...

Over at the Yale Journal on Regulation's blog, @j-p-a.bsky.social and I have put together a 🔥 symposium for you on the intersection of AI and administrative law. This symposium is for AI skeptics, AI believers, and everyone in between. Come join us! www.yalejreg.com/topic/sympos...

🚨 ATTN: Tech and admin law heads. @bridgetdooling.bsky.social and I have convened a fantastic blog symposium on AI and the APA! Follow along at Notice & Comment over the next ~week for sharp insights from a very thoughtful panel of academics and practitioners. www.yalejreg.com/topic/sympos...

A ProPublica report on plans to use AI to write regulations at the US Department of Transportation should be a warning signal for public interest advocates and litigators, writes Jordan Ascher, policy counsel at Governing for Impact. It’s time to prepare for a flood of machine-generated rules.

Last week, ProPublica reported that the DOT was set to use Gemini to write rules (or, in their words, generate "word salad"). Today, in @techpolicypress.bsky.social, I try to situate that reporting in context and dig into why it is disturbing. www.techpolicy.press/trump-admini...

NEW: The Trump administration is planning to use AI to write federal regulations despite the risk of hallucinations. “We don't even need a very good rule,” the Transportation Department’s top lawyer said of the plan, per meeting notes reviewed by ProPublica. “We want good enough.”🧵

Very glad to speak with @jessecoburn.bsky.social for this 🔥 story.

Walt Whitman:

Section 1 of the Ku Klux Klan Act of 1871—codified at 42 U.S.C. § 1983—is a cornerstone civil rights law that is more vital than ever today. Boldness like this can matter down the centuries.

It boggles the mind how much could be done for society—and how much inane case law could be abrogated—with just a few modest amendments to Section 1983.

This kind of sophistry is frequently encouraged in the legal profession. And it’s been on a roll recently!

:)

To emphasize Steve's point: Fourth Circuit suggested that employees may be able to file in district court because the MSPB no longer functions as Congress intended due to (1) a lack of quorum (fixed as of October) and (2) the likelihood that removal protections would be found unconstitutional.

Today in Lawfare: Trump's plan to screen federal job applicants for political allegiance is flagrantly unlawful. We explain why and what litigants might think about.

Reports of the death of universal preliminary relief are greatly exaggerated—at least as to final agency action. From me, today, in @justsecurity.org:

I’m encountering the word “munch” a lot lately and I do not care for that word at all. 🥴

Great & provocative paper and 🧵 from Jordan on what admin law might have to say about agencies using genAI & machine learning for rulemaking. Includes a superb research agenda for scholars wondering where to dig in. ⬇️

The Trump administration has made clear that it is going to use AI to supercharge deregulation. What do the doctrines of administrative law have to say about AI? A new Issue Brief from me and GFI explores that question. governingforimpact.org/wp-content/u...

AI is going to have a huge role in federal rulemaking. It probably already does. The public should push agencies to disclose how they are using AI. Me, in Notice & Comment, on one way to do so: www.yalejreg.com/nc/seeking-d...

I have a piece in Lawfare today exploring a supporting player in litigation against executive action—nonstatutory review. A quick take on a big topic.

"Despite the Court’s rejection of the application of this doctrine in Nuclear Regulatory Commission, nonstatutory review is likely to remain a valuable tool in pending and future challenges to the administration," writes @j-p-a.bsky.social. www.lawfaremedia.org/article/nrc-...

Extremely enjoyable baseball experience.

What comes next after CASA? “Universal” relief remains possible, explain John Lewis and Jordan Ascher. The APA and other equitable remedies can still block unlawful executive actions nationwide. A must-read primer. 👇 www.justsecurity.org/116162/unive...

Cut through the noise on universal injunctions and learn about the current state of play with me and @jtlew3.bsky.social.

The Administrative Procedure Act is all anyone's talking about these days. My org, Governing for Impact, has just put out a nine-part (!) library of primers on what APA litigators need to know. It's all here: governingforimpact.org/apa-library/

This illustrates how the Supreme Court facilitates presidential authoritarianism. By declining to enforce federal laws that bind the president—except for laws it likes—it invites presidents to challenge legal boundaries. Our democracy needs a court that enforces federal law, not one that defies it.

The Supreme Court’s notorious recent decisions construed federal laws criminalizing certain acts by state and local officials; they should not be treated as authoritative treatments of the word “corruption.”

Great bit from the MTA’s preliminary injunction motion in the congestion pricing case.

In its accidentally disclosed memo, DOJ argues that the agreement authorizing congestion pricing can lawfully be terminated under OMB grant rescission regulations. Maybe; there's a lot of pending litigation regarding those regs. But, as a first cut, the argument sounds weak to me.

“Disorder in the Court: U.S. DOT Lawyers Upload Memo Admitting Their Congestion Pricing Case is Weak” — nyc.streetsblog.org/2025/04/24/d...

But not scrubbed from our hearts or my Google Drive.

So, uh, it certainly looks like DOJ intended to file a letter to Judge Liman in MTA v. Duffy, but….it accidentally filed a letter containing its legal advice to the Department of Transportation instead storage.courtlistener.com/recap/gov.us...

It appears so.

“[T]here is considerable litigation risk in defending the Secretary’s [congestion pricing directive] under the Administrative Procedure Act, that the decision was contrary to law, pretextual, procedurally arbitrary and capricious, and violated due process.” storage.courtlistener.com/recap/gov.us...

Did the Federal DOT just accidentally file a letter containing legal advice from the DOJ, instead of a letter addressed to the Judge, in the MTA v Duffy case?