The "Pawing at Scott" Edition

Summary

Rational Security — “The Pawing at Scott” Edition

Lawfare Podcast, January 29, 2026
Host: Scott R. Anderson
Guests: Alan Rosenstein, Molly Roberts, Eric Columbus

---

Overview

In this eclectic, tech-oriented episode of Rational Security, host Scott R. Anderson is joined by Lawfare colleagues Alan Rosenstein, Molly Roberts, and Eric Columbus to dissect two major stories: the escalating Minneapolis immigration enforcement crisis and its national legal/political ramifications, plus the release of Anthropic’s “Claude Constitution,” a philosophical step in AI governance. The team weaves in local color (ice storms, unruly kittens), sharp legal and ethical analysis, and characteristic wit as they dig into some of the week’s thorniest national security debates.

---

Key Discussion Points & Insights

1. Minneapolis Immigration Crisis and National Backlash

[06:30–46:07]

The Situation on the Ground

• Context: The police killing of ICU nurse Alex Preddy by Border Patrol agents in Minneapolis has triggered a national backlash against the Trump administration’s immigration enforcement policies and ICE’s tactics.
• Alan (09:24): “The vibes are bad... Just to know that there are a mass number of masked, armed paramilitary troops in your city who are just doing a very bad job, who are not here for any obvious policy purpose... It’s a very strange place to do this because there's not a very large immigrant community here.”

Political Ramifications & Shifting Rhetoric

• Scott (16:05): Notes unusual bipartisan condemnation, including striking statements from Republicans like Chris Medell, who called out “the national Republican stated retribution on the citizens of our state.”
• Molly (17:56): Outlines political shift: “When Trump even said, when he dispensed with the language of gunmen and started to say, ‘I don’t like seeing anyone killed,’ it was as if he created permission for both those more moderate people and slightly more conservative people to also say they don’t like what’s going on here.”

Tactics, Accountability & Legal Pushback

• Leadership reshuffle: Greg Bovino, the local ICE chief, is replaced with Tom Homan, signaling a possible change in tactics but not necessarily in policy direction.
  • Molly (19:33): “He’s not as interested in just being out on the street conducting indiscriminate sweeps... But he’s also not like a softy on immigration. This is the family separation policy guy...”
• Blame game: High-level Trump advisors, like Stephen Miller, face growing criticism for their direct hand in messaging—an unusual sign of vulnerability.
  • Scott (21:41): “That’s the sign of someone on the defensive. And that particularly is important [...] because Miller is so often seen as having a direct line to the president...”

Crackdown on Protesters & Legal Overreach

• Eric (25:04): Describes highly aggressive and legally extraordinary DOJ actions in pursuing protesters—abusing the FACE Act (normally for abortion clinic disruptions) to target protesters at a church:
  • “This is an instance where they were literally making a federal case out of something that did not deserve to be one. [...] They just adopted all sorts of unusual approaches, from trying to deny bail... to seeking mandamus before the 8th Circuit...”
• Judge backlash: Chief Judge Patrick Schiltz (a Bush appointee) is “fed up” with the DOJ’s tactics.
  • Eric (31:04): “Everything DOJ is doing is out of touch with reality... the judge may respond himself... He says that everything DOJ is doing is out of touch with reality. [...] This is not some lefty squish... and he has had it, he's had enough.”

Aggressive Federalism & State Pushback

• Minnesota officials threaten criminal investigations/prosecution of federal agents; state files anti-commandeering/sovereign immunity lawsuits:
  • Alan (39:49): “Just because you, the state, allege that a federal official broke state law and they did so outside the bounds of their federal authority, you don’t have to automatically win. [...] A federal officer goes and just like murders someone in cold blood... There’s no federal immunity there.”
  • Scott (42:52): “Progressives are rediscovering federalism... There are real limits here. They've just never been litigated because no executive branch has ever pushed it this far. That’s really interesting.”

---

2. Anthropic’s Claude Constitution: Philosophy Meets AI Governance

[46:07–70:06]

What Is the Claude Constitution?

• Anthropic’s “Constitution” is a public declaration of ethical values and behavioral guidance for its flagship AI, Claude.
  • Scott (46:25): “Talk to us about what the Claude Constitution is and what it’s trying to do that’s most interesting in your mind... what it says, and then what it may be trying to do beyond that.”

“Virtue Ethics” Approach

• Alan (46:25): “When you read these documents, you’re like, ‘Huh, this feels like a philosophy article because it’s a philosophy article written by an actual reasonably professional philosopher.’”
• The system is guided by values and character traits—‘virtue ethics’—not just rigid rules or utilitarian calculations:
  • “What they’re betting on is that instead of trying to give Claude a lot of very specific rules... Instead, we’re trying to give Claude judgment, give it a list of interesting values to take into account... use your judgment.”
• Anthropic intentionally avoids micro-managing the model with endless rules, instead aiming for a more robust and flexible AI “personality.”

Practical Example & Open Questions

• Molly (53:18): “From the perspective of an everyday person who uses or thinks about AI... these machines are always going to develop some sort of persona... What Anthropic seems to be saying here is, well, let’s work on the persona instead.”
• Both Alan and Molly highlight the experiment-like quality of this approach: is “training values” more effective than writing hard rules?
  • Scott (57:11): “Virtue isn’t something you define even in vague terms on the outside. It’s actually something you encounter through practice, applied reason, and conditioning.”

Moral Status and the Nature of Machine Consciousness

• Scott (62:26): On the bold admission in Anthropic’s Constitution: “Claude’s moral status is deeply uncertain.”
  • “If we blind ourselves to the moral capacity of these actors, you really run the risk of doing something horrible... I do think there’s something there...”
• Alan (64:52): “If a chicken can have moral concern, why can’t an AI system that is already, and certainly in the future, a lot smarter than you are?... It’s not like... there are a lot of smuggled assumptions that we understand how consciousness in human beings works.”
• Anthropic’s willingness to admit the uncertainty and seriousness of this ethical problem is, for the team, both intellectually honest and essential.

---

Notable Quotes

• Alan (09:24): “The vibes are bad. I don't have much more sophisticated to say than that... there are a mass number of masked armed paramilitary troops in your city who are just doing a very bad job, who are not here for any obvious policy purpose.”
• Molly (17:56): “When Trump even said... ‘I don’t like seeing anyone killed,’ it was as if he created permission for both those more moderate people and slightly more conservative people to also say that they don’t like what’s going on here.”
• Eric (31:04): “This is not some lefty squish. This is Judge Patrick Schiltz... and he has had it, he's had enough.”
• Alan (46:25): “Turns out Aristotle was right. Or at least that’s my takeaway from this... after thousands of years of ethical debate between philosophers, we’re almost running an experiment in silico.”
• Scott (62:26): “If we blind ourselves to the moral capacity of these actors, you really run the risk of doing something horrible...”
• Alan (64:52): “If a chicken can have moral concern, why can’t an AI system that is already and certainly in the future a lot smarter than you are on some and soon to be almost every intellectual domain...”

---

Timestamps for Important Segments

• 06:30 — Start of Minneapolis crisis and national reaction
• 09:24 — Alan describes local atmosphere: “The vibes are bad...”
• 16:05 — National political shift, with unusual bipartisan criticism
• 17:56 — Molly explains rhetorical shift following Trump’s statements
• 25:04 — Eric details DOJ’s aggressive legal tactics and judicial pushback
• 39:49 — Minnesota’s state-level legal countermeasures and federalism implications
• 46:07 — Switch to AI topic: Anthropic’s Claude Constitution
• 46:25 — Alan outlines “virtue ethics” as organizational principle for Claude
• 53:18 — Molly on the inevitability & importance of AI “persona”
• 62:26 — Discussion of moral status and the ethics of machine intelligence

---

Memorable Moments

• On ice storms and unstable snow in D.C.:
  • Scott (03:04): “I kind of liken it to being a mouse walking on creme brulee.”
  • Panel: Playful pet banter, including a rogue kitten “pawing at Scott’s face.”
• Dark humor about “scalps” and Washington politics:
  • Scott (21:41): “The scalps are fine if they’re of protesters. Let’s just clarify.”
• Philosophical riffs—during the AI ethics segment, Alan and Scott trade references to Aristotle, rolling Dungeons & Dragons jokes, and child-rearing metaphors.

---

Object Lessons (Recommendations)

[70:34–76:47]

• Eric: The “Tingler,” a head-massage device, for migraine relief.
• Alan: Season two of The Night Manager (spy thriller with Tom Hiddleston).
• Scott: “Fire Snake” method for de-icing D.C. Metro rails and the joys of good long underwear in winter.
• Molly: A memeable, Mandarin-embroidered “century of American humiliation” hat as a satirical commentary on global politics and pop culture.

---

Overall Tone:
The episode balances seriousness regarding legal, political, and ethical crises with the hosts’ signature warmth, wry humor, and intellectual curiosity. Both the Minneapolis segment and the AI ethics discussion are accessible yet thoughtful—anchored with personal anecdotes and a sense of the surreal pace of change in 2026.

---

For more details and deep dives:
See Lawfare’s show notes and Alan’s referenced Scaling Laws podcast episode on Claude’s Constitution.

Transcript

[A]

Your new home is now ready. Dr. Horton, America's builder, has new homes that are ready today. With new construction communities in Ellensburg and throughout the Greater Seattle area. Dr. Horton has the right home for you. At Dr. Horton, we're still building with flexible living spaces, smart home technology and two and three car garages. More communities and more homes available every day. Find your new home in Ellensburg now [e-mail] Dr. Horton, America's builder and Equal Housing Opportunity Builder hey folks. Scott R. Andersen Here it is cold here in our nation's capital. There's frost on the grass, ice on the sidewalk, and a small pile of hats and gloves sitting by my front door for whenever the family and I store up the fortitude to step outside. It's the time of year when you really need your winter wardrobe to deliver for you, helping you stay warm on the inside while looking sharp on the outside. Luckily, Quince has everything you need, from their amazing Mongolian cashmere sweaters to wool coats and other outerwear that really lives up to the daily grind while still looking good. It's not just coats and sweaters. They've got hats, gloves, socks, scarves, everything you need to help keep those winter winds at bay. Or if you're staying at home, and especially if your heating system is like mine, a bit on the old and temperamental side, curl up in some of Quince's fantastic bedding, Wear some of their phenomenal blankets and quilts, maybe even on top of a performance velvet Quince couch while you're at it. Personally, I have been absolutely living in one of Quince's stretch sweater fleece shirts this month. On the outside, it looks like a rugged plaid that suits my bearded aesthetic, but in practice, it's more soft, warm and cuddly than you'd ever expect. Just like me. To refresh your winter wardrobe with Quince, go to quince.comsecurity for free shipping on your order and 365 day returns. Now available in Canada too. That's Q U-I-N C-E.com Security free shipping and 365 day returns. Quints.com Security now let's get back to the show. Well, I'm glad you all appear to have survived the coming of winter, of true winter. Here Alan is smugly smiling into his mug, his Minnesota mug, as he slowly sips a hot beverage and laughs at our incompetence here at an Asians Gap along with most of the country. But I say this is the weirdest snow I think I've ever Encountered. And I can't quite think of a parallel. Like. Alan, do you get this? Do you get this, like just solid ice snow in Minnesota? Is this a common occurrence?

[B]

Well, so what? I don't have a sense of. So what, what are you guys dealing with? Because, because for us, we've had. For us the issue has been temperature. So we got to like -20 on Friday, which was bad. That's when they closed the schools. That's not great. But we actually haven't had any precipitation. And, you know, other than the frigid temperatures, it's been like really lovely here. It's been, you know, bright blue skies and stuff like that. So what is the, what is the snow situation in our nation's capital?

[C]

Well, part of the situation is it's not really snow, it's ice. I think that was the issue. It snowed a few inches and then it sleeted and then it got cold again. So the snow iced over and it's like being on a glacier. It's crazy.

[D]

I kind of liken it to being a mouse walking on creme brulee.

[B]

It's not as tasty. I mean, if you're the mouse, you can eat the crunchy topping and unfortunately that is not tasty.

[D]

But you can also fall through it and from step to step, you don't know which one it's going to be. So it keeps it interesting.

[B]

Yeah.

[A]

I had to drive my daughter to daycare to start daycare because affiliated with a hospital system was miraculously open, one of very few in the city. And it was perilous. It took me like, I think daycare is half a mile away. It took me like 30 to 40 minutes round trip because I had to get my car in a 30 point turn out of my garage after having shoveled it all out last night. But because the ice was just so piled up everywhere, there's not enough place to put all the ice and snow because none of it's melting. So I just don't know how exactly. I mean, I'm not gonna lie. My experience driving around today was like, yeah, we don't want people driving around in this. People should just stay home. But like, I don't see, I don't see when that changes because there's nothing you can do with all this ice.

[C]

A kitten showed up at my home on Friday right before the storm, and we had to take it inside because the giant storm was coming. But now we just have this kitten. We can't. We have found somewhere to adopt it.

[A]

Certainly indefinitely at this point, I think that's the kitten rule. Like once you're stuck with it for three days, you kind of own it at that point, it's hard to get rid of it.

[C]

That's what people say. People say, oh, this cat adopted you. There's some discord in the household as to whether that's true. And we can't get it anywhere else. We can't drive. So it's just here. It's pawing at the screen right now. It's pawing at your face right now, Scott.

[A]

I'm deathly allergic to cats, so please don't back the fuck off. That would be great.

[D]

I think listeners often are pawing at their screen when Scott is.

[A]

Yeah, exactly. Uncommon occurrence.

[B]

It's a range of reasons why they're pawing. Some positive, some not so positive, so negative.

[A]

Hello, everyone, and welcome back to Rational Security, the show where we invite you to join members of the lawfare team as we try to make sense of the week's big national security news stories. I am your host, Scott R. Andersen. Thrilled to be back again this week with some of my teammates as we sort through some of the headlines. We're going a little more eclectic, a little more tech oriented on the headlines because I'm joined this time with by co host emeritus and lawfare. I say senior editor but not senior editor. Research director. Research director. Now Professor Alan Rosenstein. Alan, thank you for coming back on the podcast.

[B]

My pleasure. Even through professor there. My goodness, Scott. We interact a dozen times a week. I feel like we're still friends. You don't have to be smart.

[A]

I know, but you're dressed like a complete slob, so I feel like I have to clarify why you're on the podcast. It's just part of the game. No, just kidding. It's a casual snow day today. I think we can all relax a little bit. I certainly am dressed down a little more than I might be on other office days because I haven't been out of the house in a week. At this point, I'm going slightly insane along with my children. And joining me on this journey through sanity as well, Art Lovers 2. Not quite newest, I guess. Second and third newest at this point, senior editors, Molly Roberts. Molly, thank you for coming back on the podcast.

[C]

Yeah, happy to be here always.

[A]

Even though you were sicking your feral felines on me. Evidently in digital form as we speak. But that's okay. I won't hold that against you. And Eric, I don't know what. Eric Columbus, our other senior editor. Thank you for joining us as well. I don't know what sort of livestock or animals you have around the house that stick at me, but I know you have a couple around the house, so we'll see how they react to my general presence. Oh, there we go. Right next to you. Yeah, that guy's not sicken on anybody. He's, he's a sleepy boy. That's okay.

[C]

I've got one on the floor doing that too. It's really the duality of pets here. The kitten is going insane and the dog is fast asleep and a complete angel.

[A]

Oh, I love my, my poor dog because like she does not like the ice and the snow. Like we try and take her out and she like turns right back around about 10ft out the door. So she's going stir crazy. She spent, she's like a six year old dog, spent maybe like 30 minutes just doing loops around our small row house living room last night. So I'm impressed you're able to get your dogs this settled down during daylight hours. Regardless, we don't actually need to talk about this type of ice. We have other types of ice to talk about, among other issues for this week. So let us get down to brass tacks for our first topic this week, slipping down the slope. Last Week's killing of 37 year old ICU nurse Alex Preddy by Border Patrol agents in Minneapolis, Minnesota has triggered what increasingly appears to be a national backlash against the Trump administration's immigration policies and isis. Violent tact. Republicans and Democrats alike have been increasingly public in their criticism of the administration's actions and in particular DHS Secretary Kristi Noem, while state officials have begun exploring more legal avenues for pushing back against federal actions. The Trump administration, meanwhile, may be shifting tack as it has replaced Border Patrol commander at large Greg Bovino with immigration czar Tom Homan on the ground in Minneapolis and begun to adopt a slightly more conciliatory tone, at least in certain dimensions. Is this a real turning point for the Trump administration's flagship policy or more of a feint? Topic 2 Now we're just waiting on artificial strength, dexterity, wisdom and charisma. Last week, the artificial intelligence company Anthropic.

[B]

Oh, I'm sorry, Scott. That was so good.

[A]

It's a pretty good joke. I've actually used that joke once before, but it was good enough to recycle, I think, six or nine months later. So I'm just digging back and we're going to knock those off one by one. I get to use the joke every time we get through another one. But last week, the artificial intelligence company Anthropic released what it is calling its Constitution for its premier AI model, Claude. The Constitution seeks to instill a moral framework, value system, and even personality into the AI model, taking an unprecedented step in both private AI governance and, to some extent, AI personhood. How does Claude's Constitution factor into broader discussions about AI development and regulating, how models should interact with users? So our first topic we are in the rare position to go to an on the ground reporter of some sort, and that is you, Alan. You are of course in Minnesota, in Minneapolis. I actually don't know exactly where you live, but in the vicinity of Minneapolis, if not in Minneapolis property. It has been, I have no doubt, an incredibly intense several weeks for everyone in the community in Minneapolis. Even if you haven't been as directly affected as many others there about what's been going on. I'm sure that's got to be particularly true around the law school, given how many of these issues center on questions of the law, state and federal separation of power, civil rights, civil liberties. So before we dig into some of the substance of the update for this week, I think this is one of those stories that frankly, we're going to be talking about almost every week as long as it lasts, because there's always new issues that keep popping up every day. Talk to us about a little bit what it's been like on the ground about the sentiments in the state, in the city, and the trajectory those have been on over the last several weeks of really just every day, just like dramatic, shocking national news making developments.

[B]

Yeah, the vibes are bad. I don't have much more sophisticated to say than that, but it is this feeling of kind of oppressiveness of a city. And you're right, I'm not in Minneapolis. I'm in St. Paul. And I'm going to piss off all our Minnesota listeners by saying this. It's basically just one city for all intents and purposes, and I'm sort of right in the middle of it. The vibe here is just very, just very sort of depressing. You know, look, I should say I don't want to steal any valor here, right? I mean, I'm not and I am not an undocumented sort of. I'm not sort of here without authorization. So I'm not affected in that way. I don't know a lot of people personally who are affected that way. I have not been in the streets being pepper sprayed or shot. So lots of people are experiencing this much worse than I am. But Just to know that there are a mass number of masked armed paramilitary troops in your city who are just doing a very bad job, who are not here for any obvious policy purpose. Whatever you think about the question of illegal immigration and the appropriateness of deporting people who are here, Minneapolis and Minnesota, it's a very strange place to do this because there's not a very large immigrant community here. The only border that's close is Canada, and other than moose and geese, there's not a lot of people crossing that. The main sort of immigrant community here are Somalis, but the vast majority of them are actually citizens. So I'm not saying that this kind of behavior would be appropriate near the southern border, but if the goal was actually to find a lot of people who are here illegally, especially those who are committing crimes and deport them, this is just a very strange place to do that. It's pretty clear that the reason that Minnesota is being targeted is some combination of anti Somali bigotry, a feeling that the legitimate reportage and concerns about the welfare fraud that has been covered extensively over the last few years, that that creates sort of a political opportunity for the administration here. And I think just anti Waltz animus by Trump because Waltz was the Vice President and said vice presidential candidate, obviously sorry, I should say, for Kamala Harris and said mean things about Trump and Vance. You know, all of this is creating just frankly pretextual reasons for a level of repressive law enforcement and frankly, very poorly trained and poorly executed law enforcement. All of that is a, a remarkable thing to realize that one is living through in 2026. And you mentioned this question about the law school. Again, this has not, I think, affected the law school directly, as far as I can tell, in any particular ways, though of course there may be scenarios that I'm not personally privy to. But I'm a law professor, I'm a con law professor, I'm a Krimpro professor. And it is quite sobering to reflect on the almost law free nature of a lot of this. And again, I'm not myself a particular radical about immigration policy. Right. I mean, I do not take the position that there's no such thing as an illegal immigrant or that deportation is in all cases unlawful.

[A]

Right.

[B]

I mean, we have immigration policies and we can have a whole range of discussions about how those should be structured and how they should operate. But I would say to those, especially on the right, who have suddenly discovered a great love of overwhelming government enforcement of the laws, how would you feel if the AOC administration sent tens of thousands of poorly trained paramilitary IRS enforcers into your city to enforce the tax law. I think you can appreciate the similar sentiment here. A large number of the law school faculty signed an open letter which is floating around expressing our concern with what's happening. It was quite a cross ideological group of law professors and I was part of that group, despite having as a rule for myself over the last 10 years of being in the academy of never signing on to open letters because I think the entire genre is almost never justified and is almost always a bad deal. But I signed onto this thing pretty quickly. Just reflecting that at some point the situation on the ground becomes so grotesque from a perspective of the normal ordering of legal events that I felt I should say something. So I apologize. That's somewhat long winded and rambly way of saying it's bad. And I will say I remember Scott back in RATSEC version 2 when it was Yumi and Quinta and I was the sort of pathological, both sides moderate.

[A]

Now I have to do it to myself.

[B]

Exactly. I will say I have completely lost any sense of chill over the last year and especially over the last few weeks. And the idea that we could have these sorts of half incompetent, half vicious paramilitary operations throughout this country strikes me with sort of immense dread. And I'm not saying I wish it happened anywhere else, but it's particularly painful to see this happen to me, my city.

[A]

It's really remarkable and thank you for sharing that. Alan. Living through these sorts of moments, even at a little bit of a distance is never easy. But I think what you're describing reflects something that I feel like I've seen more and more particularly this past week since the really horrendous killing of Alex Pettis. In some ways, I think amplifying and doubling down on the fact that there were nay Goode killing about two weeks ago, a little more than two weeks ago now, was not an isolated incident and it's part of a trend is that you are seeing these sentiments much more vocally from a lot of corners that you didn't for a long time. A lot of corners that are not inherently even hostile to the Trump administration's general immigration policy. A lot of Republican corners as well. I mean here I think one of the incidents I thought was really compelling and interesting is the case of Chris Medell. I may be saying that wrong. Minneapolis attorney who was a gubernatorial candidate for the Republicans, participated in debates relatively recently. I don't think he was considered a frontrunner by any stretch of the imagination. But he ultimately withdrew and released this pretty compelling, powerful statement, frankly, because of the rhetoric used, basically said, let's see if I can find the exact quote here. I cannot support the national Republican stated retribution on the citizens of our state, nor can I count myself a member of a party that would do so. And also calls out the racial discrimination that's happening against people of different racial minorities in Minneapolis. It's really, really compelling and a notable statement from somebody who is aspiring to be a leader of a party in this city at this particular moment. And we're seeing nothing, I think, quite that striking or strong. But you hear pretty targeted criticism from people like Thom Tillis coming from North Carolina, from Lisa Murkowski. Now, these are the people who are always more vocal in criticism.

[B]

I hear Susan Collins is very concerned.

[A]

Scott, Very, very concerned, Twilling her thumbs nervously as the state of Maine appears to be next on the target list.

[B]

Enjoy that, Susie.

[A]

So, Molly, let me come to you on this and let's talk a little bit about, like, the national ramifications of this, because in the last week we've seen a kind of interesting bit of back and forth, right. We saw this effort to spin the Pettis killing as a very strong rhetorical move almost immediately by the Trump administration, casting it as he was an attacker intent on massacring the agents because he was armed. Quickly, video evidence came out showing that in fact, he had been disarmed before the shooting by the officers. And that wasn't really doing anything threatening. It was filming officers and appeared to try and help a woman who had been pepper sprayed or sprayed with something. I think it was pepper spray when she fell and get kind of between them and the officers. And that was attained as part of that. You saw that narrative fall apart really almost immediately, trigger a lot of backlash from a lot of different corners, including corners of the Internet, traditionally very friendly to Trump country. I've heard it on sports podcasts and radio talking about this incident, which is really compelling as it's fairly nonpartisan, right leaning. Part of my media ecosphere probably more than anything. So talk to us about what the political dynamics have become around this incident and what it's translating into, as far as you can tell, for the broader policy dynamics around this whole issue set for the Trump administration.

[C]

Yeah, I think it was an interesting phenomenon where at first you had people like Susan Collins being very concerned, Lisa Murkowski speaking out a little more, frankly, robustly. And that's familiar that you have those people pushing back and then you had people who were maybe a little less on the moderate side pushing back, too. But then when Trump even said, when he dispensed with the language of gunmen and started to say, I don't like seeing anyone killed, it was as if he created permission for both those more moderate people and slightly more conservative people to also say that they don't like what's going on here. People who'd maybe been holding back because they were worried that there'd be backlash from Trump on that. So I think that's definitely interesting. Whether there's going to be any big change in policy, I don't know. I think that the only concrete action we've seen so far is getting Greg Bovino out of there and replacing him with Tom Homan. And on the one hand, that might be meaningful from a practical policy perspective, because Tom Homan is more interested in targeted raids. He's not as interested in just being out on the street conducting kind of indiscriminate sweeps. But. But he's also not like a softy on immigration. This is the family separation policy guy. And, you know, this is the guy who took a $50,000 bribe in a kava bag. So, you know, I don't care that much of Obama gave him a presidential rank award or whatever it was a few years ago. Like Caroline Levitt was boasting about, at.

[A]

Least have the dignity to do it in a sweet green bag. You know, I feel like that's more dignified, kind of the more DC Elite set, if you're really going to play that game.

[C]

Well, I was always, I was kind of thinking, you know, you don't even need to. If you want to bribe me, just give me kava in the kava bag. And I'm happy you heard it here first, friends.

[B]

I hear you send Molly kava gift certificates.

[C]

But in any case, yeah, he's not some, you know, big bipartisan immigration figure. So I think we already knew that the polling on precisely how ICE has been acting was bad. Americans don't like it. Some Republicans are probably relieved that Trump is sort of starting to suggest now that he doesn't like it either, because they don't want to have to cross him by saying that they don't like it. So. So I can see maybe a change in the degree of aggression. I think they're certainly trying to do more command and control from the top and make sure that there isn't this really obvious rampant lawlessness. Although I suppose it depends on your definition of lawlessness. They don't want executions of Americans on the streets. But they still probably do want to be pulling people over and they certainly do want to be conducting raids and they do want to be doing deportations at a greater scale than maybe your average American wants. And we saw that. We saw that. I believe yesterday there was data on how many arrests there had been in Minneapolis, and it was no fewer than there had been prior to this change in rhetoric.

[A]

Yeah, it's one of these things we're going to have to keep an eye on to understand what the macro trend is. I mean, the thing that strikes me that's interesting. Interesting is the way this maybe disrupts the decision making practice or process or people involved on some of this stuff, or seems like it might have, because this is clearly a case where there are lines of blame being drawn to not just Bovino, which is the first scalp we've seen to get cut off of this, despite the administration's very vocal, expressly described no scalps rule that they've had since the beginning of this administration, which has been. Mike Waltz was kind of like a half exception to that. He got promoted to a different job. That's less important. Right?

[B]

To be clear. To be clear, go. The scalps are fine if they're of protesters. Let's just clarify. So no internal scalps rule.

[A]

And Bovino is not one of Trump's guys necessarily. He was more or less a career guy who kind of rised to the role. So it's not clear how this fits with that sort of policy. He's clearly somebody who is being shoved, thrown under the bus for at least part of this effort. You've got Kristi Noem being targeted by a lot of folks in the Senate and other places. And then we have Stephen Miller getting direct lines of criticism, not least because he really does appear to have been the person who framed and really amplified that Pettis narrative very early on. And you saw some of the initial efforts to frame initial descriptions of his role in that decision making process get reframed again and again and again by his wife on our podcast through Mark Caputo. A reporter who's kind of had one of these stories has been a bit of a moving target over the last few days. Exactly what Miller's involvement has been. That's the sign of someone on the defensive. And that particularly is important to me in my mind, because Miller is so often, as we saw in the WhatsApp chat signal that Jeffrey Goldberg found his way into a little shy of a year ago now, he is a guy who's seen as having a direct line to the president being able to speak for the president in a way that very few people in the White House can. Maybe this begins to disrupt in that sort of line. But that's not a policy shift. Maybe that's a shift in tactics. I'm not sure it changes the overall dimension. Then again, the tactics are a big part of the problem here. That is what's killing Americans. So that's not insignificant. But I'm not sure it changes the overall policy dynamic.

[C]

Right. I mean, I think maybe more discipline on the streets. But even if you look at the discussions that Trump had with Tim Walls, with Jacob Frey, what they said they pushed for in those discussions were concrete changes. Waltz was focusing in particular on independent investigation. Frey was focusing more on don't have ICE trying to play this role of being out on the streets of the city and all day long. And none of those changes have happened. And you saw Trump kind of come back a little bit at Jacob Frey today when he said that Minneapolis wasn't going to enforce federal immigration laws because, you know, Jacob Frey doesn't want ice. They're enforcing federal immigration laws, but he doesn't want Minneapolis to enforce federal immigration laws. So how are you going to get past that impasse with Trump? Trump is clearly upset about it. He says he's playing with fire. So we'll see.

[A]

Yeah, not going to help the commandeering lawsuit necessarily. Anti commandeering defense for the administration. That's all right, Eric, let me bring you in here. I know you've been looking into, you know, there's a flip side of this, which is that Trump administration, even it's ameliorated its rhetoric. This is coming on the heels of some of its most aggressive legal action we've seen. And particularly we have the case where a number of protesters went to a church to protest. I think it was an ICE personnel who was the pastor of this church. And we've seen an aggressive effort to prosecute them. Prosecute journalist Don Lemonade, who was involved in kind of interacting with them before the protest. Talk to a little about some of the work you've been doing on that case and what it tells us about the tactics the Trump administration seems to be trying to pursue on the continuing aggressive front through the legal front of the Justice Department, where Pam Bondi hasn't as visibly walked back any sort of rhetoric that she's been issuing throughout this campaign about coming down hard on people not obeying the law and where that fits into the broader legal picture. We're also seeing pushback from the other side from state officials now as well. So square that circle a little bit for us.

[D]

Sure. I'm working on a piece with a couple of colleagues at Lawfare on this case where they've been trying the administrative. There were protesters who rather rudely, at the very least, disrupted a church service on, I believe, I think, two Sundays ago. And it became kind of a big deal on the right was Don Lemon was live streaming it. And so it got a lot of immediate attention out there on the Internet. And within hours, Pam Bondi and Harmeet Dhillon, who's the extremely online head of the Civil Rights Division, were saying that they will basically make people hang for it. I'm paraphrasing. But they said they would bring the people to justice who did this. And this is an instance where they were literally making a federal case out of something that did not deserve to be one. It's obviously, it may be a local Minnesota trespassing case at most, or maybe there are laws about disrupting religious services in Minnesota that might apply, but there is really nothing on the. It doesn't seem to do that type of thing that warrants a federal action. And what she was trying to use was something called the Freedom of Access to Clinic Entrances act, or the FACE act, which most people, including me, had always assumed only had to do with, mainly with abortion providers. As you could tell from its name, access to clinic entrances. A church is not a clinic, typically. And. But there is. It turns out that through a bit of political horse trading back in the early 90s, the act also applies to people who impede religious services, who impede, actually on the same terms as abortion clinics, people who impede either access to or egress from church services. They wound up successfully indicting only three and on conspiracy charges, not directly under the FACE act itself. And in the process of doing this, they just adopted all sorts of unusual approaches, from trying to deny bail to three of the people that they had successfully got indictments for on the theory that they were risk to the community or that they might pose a flight risk. And these are people who were. One was a local activist and one was a board member of, I think, the local Board of education in St. Paul. I may be getting that somewhat wrong, but she had some office in St. Paul, and the government noted that as part of its reason why bail should be denied was that they were in hotels and local hotels, they left their residences, and they had done so because of the heated rhetoric from Department of Justice had caused them to get all sorts of death Threats and pushed them out of their houses into hotels. So it's kind of ironic that they basically are. They have created the threat against people, and then they're using those threats as a pretext for trying to deny bail to them. Unsuccessfully, I might add. They also really pissed off the chief judge of the district court in Minneapolis with this case by they were trying to appeal the no probable cause decision of the magistrate judge as to 5 of the people on both charges that they sought and the Face act charge for the other three. And they appealed the determination to the district court, which apparently is unheard of, at least within the 8th Circuit. The judge said he canvassed his colleagues in other district courts within the 8th Circuit. No one could find an example of DOJ doing this, because typically there's no reason to appeal an arrest warrant application, because if you're denied it by a magistrate judge, it means that your case has a problem and you're not going to be able to convince a jury beyond a reasonable doubt. So you would either one, give up, or two, try to bolster what you have by submitting more facts into the warrant, into the arrest warrant application, or alternatively taking your case to a grand jury, which is another way to get an arrest warrant. But they were insistent, apparently, that these arrest people had to be arrested before Sunday, because otherwise there would be more disturbances, they said, at this church. And so the judge was like, look, I don't want to deal with this now. I want to wait until Tuesday and talk to my colleagues. And another kind of ironic episode of the events kind of leading to weird consequences. He was supposed to talk with his colleagues on Friday, his other colleagues on the district court bench in Minneapolis, but that had to be delayed because of security concerns in the courthouse, owing to the appearances, the initial appearances of the people who were arrested in this case, plus the visits of J.D. vance and Pam Bondi to Minneapolis to bang the drum in favor of the administration's immigration efforts. And the DOJ was not happy with that. So they took the extraordinary step of seeking mandamus before the 8th Circuit. And mandamus is a means in Latin, we order, basically or we command.

[A]

And it's a pretty extraordinary remedy for folks who have not dealt with appellate. In district courts, it's about the most demanding remedy you can try and get for district court.

[D]

Yeah. And they sought a writ of mandamus, which is basically, you're basically suing the judge in such a case. And typically, whoever's on the other side of the case basically files an opposing brief supporting whatever the judge did, or more often, whatever the judge refused to do. But here there was no one on the other side because the whole point was that the magistrate judge refused to issue arrest warrants, refused to bring these, these normal citizens into the judicial process. And so in, in that case, the, the judge may respond by himself. But unfortunately the DOJ didn't actually tell the judge that they were, were doing this and they didn't give him. You're supposed to, under the federal rules, give a copy of the action to the judge. So instead he gets an email from the AD Circuit on Friday as he's working at home because he is a mentally disabled adult son whose program was closed because of the snow, rather because of the frigid temperatures that Alan referred to earlier. And so he's working home. He gets an email from the 8th Circuit clerk saying that this mandamus petition has been filed against you and you need to respond in two hours. So it's kind of amazing. He sits down and writes a long letter to the chief judge of the 8th Circuit saying, Look, you know, I've gotten no notice of this, but I can pretty much give you a guess about what it's about. I think I know what's going on. And he goes at great length for four pages about why this is DOJ's action is unprecedented. He submits that and eventually gets. Eventually the 8th Circuit figures out what's going on and sends him a copy of what they submitted. And then he writes another letter explaining, he says, I've given you an hour to write this letter, but I'll just do it and turn it around. He says that everything DOJ is doing is out of touch with reality. He says, I apologize for responding in the form of an email. I'm also dealing with a number of emergencies, including a lockdown at the courthouse because of protest activity, the defiance of several court orders by ICE and the illegal detention of many detainees by ice, including yesterday a two year old. Now, I should note this is not some lefty squish. This is the Judge Patrick Schiltz appointed by George W. Bush in 2006, a two time Scalia clerk on the D.C. circuit and the Supreme Court. And he has had it, he's had enough.

[A]

It's a really extraordinary exchange, it really is. And I feel like it's representative of this kind of broader degree to which you're really seeing this brinksmanship by the Justice Department encounter a degree of skepticism, whether because of somewhat faulty lawyering, because of the extraordinary arguments, because of extraordinary representations, how does that affect the bottom line of the strategy? I'm coming to you. I mean, is this unsustainable from their perspective, or is this just a sign that. That they care about is getting these initial arrests having the perp walk? We saw them digitally alter the image of one of these protesters after they were arrested and posting on the White House social media. What is the bottom line that we should take away from? I mean, is this a sign that these policy strategies are failing or that they are getting what they want out of them, which just isn't criminal prosecution?

[D]

That's an interesting question. I mean, a lot of this I mentioned about the Harmeet Dhillon being extremely online. This administration is extremely online, and it's, it cares at some point, I'd say more about attention than about results. Now, that's not true about, say, Stephen Miller. Stephen Miller, I think, really deeply believes in what he is doing, just as much that I deeply believe that he is wrong to be doing what he is doing. And ironically, if there were a way to. To if this were kind of causing actual. If stepping on rakes like this were actually harming their ability to get things done in immigration enforcement, you might see someone like Stephen Miller try to push back and say, hey, guys, let's recalibrate our strategy a bit. But it's not clear, frankly, that it is because they push and push and push and they take what they can get. And if they get reversed, if they lose, then maybe they comply and maybe they don't comply, as we've seen in some of that, the language from Judge Schiltz that I read. But they don't necessarily have. They're not necessarily incurring huge costs here. And they also know that the Supreme Court is, in this case also, the 8th Circuit is more generally more favorable to them than district court judges, so they may not have much fear of pissing off district court judges.

[A]

Hey, folks. Scott R. Andersen here. It is cold here in our nation's capital. There's frost on the grass, ice on the sidewalk. Locked in a small pile of hats and gloves sitting by my front door for whenever the family and I store up the fortitude to step outside. It's the time of year when you really need your winter wardrobe to deliver for you, helping you stay warm on the inside while looking sharp on the outside. Luckily, Quint's has everything you need, from their amazing Mongolian cashmere sweaters to wool coats and other outerwear that really lives up to the daily grind while still looking good. It's not just coats and sweaters. They've got hats, gloves, socks, scarf. Everything you need to help keep those winter winds at bay. Or if you're staying at home, and especially if your heating system is like mine, a bit on the old and temperamental side. Curl up in some of Quince's fantastic bedding or some of their phenomenal blankets and quilts. Maybe even on top of a performance velvet Quince couch while you're at it. Personally, I have been absolutely living in one of Quince's stretch sweater fleece shirts this month. On the outside, it looks like a rugged plaid that suits my bearded aesthetic, but in practice it's more soft, warm and cuddly than you'd ever expect. Just like me. To refresh your winter wardrobe with quints, go to quints.com security for free shipping on your order and 365 day returns. Now available in Canada too. That's Q U-I-N C E.com security free shipping and 365 day returns. Quince.com security now let's get back to the show.

[B]

Not every sale happens at the register before AT&T business Wireless checking out customers on our mobile POS system took too long. Basically a staring contest where everyone loses. It's crazy what people will say during an awkward silence. Now transactions are done before the silence takes hold. That means I can focus on the task at hand and make an extra sale or two. Sometimes I do miss the bonding time.

[A]

Sometimes AT&T business Wireless connecting changes everything.

[B]

Hey listeners. Meet Russell.

[A]

Hey.

[B]

Russell just launched a fitness app and.

[A]

He needed to get the word out.

[B]

To busy professionals looking to stay. They fit.

[A]

So I turned to acast. I used their Smart Recommendations feature to easily find shows that talk about health and fitness. Booking sponsorships through their platform was a breeze. And just like that, my app was in their ears during their morning run.

[B]

Sounds like a smart move. Russell.

[A]

How's business looking now? Sweat is pouring and so are the installs.

[B]

Spread the word about your business with podcast ads on acast. Start today at Go acast.

[A]

So there's a flip side of this equation as well, which is that we are beginning to see state authorities, or some cases, advocacy groups and other people begin to mount up other sorts of legal attacks on some of what the Trump administration is doing. In particular, we've gotten reports, I think just over the last few days about city, county, state officials and prosecutors potentially looking into doing criminal investigations or prosecution of certain federal officials involved. Notably that particularly around the pretty killing where we saw federal agents actually confiscate A bunch of the evidence before it could be collected or analyzed by local officials who are actually in charge of murder investigations and other killings. At least a matter of first instance. So an issue there, one that actually could raise other state criminal law questions, I suspect. And then we have a lawsuit from the state of Minnesota itself challenging what the Trump administration is doing on a variety of essentially sovereign immunity principles, 10th amendment and a commandeering principle. Alan, you are a common law professor. This is kind of to some extent core federalism questions. You and I have talked about this before on this podcast as well, in the context of the California kind of masking law. Talk to us about how you're viewing these efforts in this particular moment.

[B]

Yeah, I mean, that's why I wanted to jump in on this because I don't remember when this was. Maybe a month ago or two months ago. And you and I think got pretty well into it. I was quite skeptical of this broad based California anti masking law. And you thought there might be some legs there and I don't know, we had a fun back and forth on that. I think this is very, very different. This is not, as far as I can tell, the state of Minnesota purporting to invalidate ICE enforcement in Minnesota. What they are saying is if we think federal officers have violated federal law, or at the very least if they have clearly operated outside the boundaries of federal law law in a way that violates state law, though frankly, murder, or at the very least negligent homicide during a federal operation is quite possibly a federal offense in and of itself. If we have found that these people have done this, then we will try to prosecute them. Which by the way, that is not, I think, a particularly controversial statement. Right. There have been a lot of historical pedigree. Now, of course you have to do this on a case by case basis. Just because you, the state, state allege that a federal official broke state law and they did so outside the bounds of their federal authority. You don't have to automatically win. Often those cases get removed to federal court and then a federal judge will decide whether or not the federal officer was acting within the scope of their authority and the supremacy clause, except blah, blah, blah. Right. But you know, a federal officer like out on a, you know, I forget what the, what the legal term is. A frolic.

[D]

Frolic and detour.

[B]

Thank you. Like out on a frolic and detour goes and just like murders someone in cold blood, which again, I'm not saying this is what happened in these cases necessarily, but you can imagine A case where that is, there's no federal immunity there. There's certainly nothing called absolute immunity or whatever nonsense. J.D. vance, a graduate of the esteemed the Yale Law School, seems to think that there is. So I think it's a perfectly reasonable thing to do. Whether that will work again depends enormously. And I think at this level we're operating as much at the level of sort of high stakes politics with the White House as we are as a legal domain, but that's where we are. As to the anti commandeering, this is the idea that although a state government cannot interfere with federal activities, the federal government cannot require the state government to effectuate federal law. So as much as Trump may be pissed that Minneapolis Mayor Frey doesn't want to collaborate with ice, he has no power to punish the state from, from doing so. Now we can have a conversation about whether Congress can make certain funds contingent on this, that and the other. That's kind of an interesting question. But you know, certainly given everything that's going on, there's nothing facially unreasonable about these legal moves. Again, whether they will work out ends up being just a very, it's a very case specific fact bound determination.

[A]

It is really, really interesting and from a strictly like removed academic perspective, like fascinating new zone of, of constitutional law that's going to get litigated if these things ultimately be brought to court. I think is really, really interesting.

[B]

Progressives are rediscovering federalism.

[A]

Yeah, absolutely. I mean, and there's a reason why a lot of people liked federalism over the years, right? In various ways. It has a lot of problems. It can cause a lot of problems too. But there are upsides of it and it is a particular case. I mean, in the case that is, I think cited most commonly regarded as the Inry Nagel case, it actually ironically is like the case that is also the progenitor of the protective power or one of the main, I think, I think, pillars of the protective power principle that the Trump administration has invoked for deploying military forces in California and Portland and Chicago, where basically you had a federal officer kill someone who tried to assault a federal judge, if I recall correctly. And there's a question of was this within the scope of their official duties? And the court came fairly broadly and said, well, yeah, it's kind of implied by statute, implied by executive power. You have to be able to defend federal agencies as part of the federal function. And the language, if I recall correctly, that they use is kind of whether it's necessary and proper, which kind of makes sense. I don't think they actually directly tied to the constitutional provision, if I recall, but it kind of makes sense because you're necessary and proper clause. That's the furthest extent of the federal government's legislative authority under the Constitution, which is limited. The rest falls to the states. So there are real limits here. They've just never been litigated because no executive branch has ever pushed it this far. That's really interesting. And the assumption that you get when you hear J.D. vance talk about absolute immunity, it's because A, the Justice Department often frames it as near absolute immunity because it's so broad and Federal authority is XYZ super broad. This would have Republicans from 30 years ago rolling over in their graves. But regardless, that is kind of the line of attack that's taken. And B, a lot of the court judgments have never wrestled with these cases. The government's always won, but it's not a randomly distributed sampling of the potential zone of the law. It's because the executive branch is being incrementalist in these things and now you're going way out there. I think the executive branch Trump administration has to be way more worried about this than they are or at least they should be way more worried about this because I think there is a lot of these cases where they're really pushing in a domain where the link to federal authority is really limited and that gets to be hard. And when you already see judges being as skeptical as them, as we've already seen, as Eric's described for us in other contexts, same judges are going to hear these cases even if you remove it. Inevitably if you try and prosecute a state level official, they're going to remove it to federal court. Then it's going to be the same district court judges. It's the same kind of factual assertions, assertions about the directions they're given, the measures taken for accountability for oversight, that sort of thing that's going to bear on was this person operating within the scope of their duties in a reasonable way and burning your credibility with federal judges hurts you on those sorts of assessments. So long story short, it's perilous, I think legal terrain for the administration to really be betting on. But we'll have to wait and see. I could be wrong about that. Let us go as we are well past our usual time on this segment, let us go from the US Constitution to the US Constitution, and by us, I mean we humans and of course our robot overlords of at least one stripe, I.e. claude, the AI model for anthropic, we had them drop a massive document that I have not spent the time to read quite yet. Although I've perused it, I've read a couple of great pieces on it, including, I will say, Alan, you're a really phenomenal piece for Lafaran. That's what I think was short, succinct, but really pulled out a thing. Reminded me of those crazy, heady days of college where people would stay up late and get a little crazy, start experimenting with Nicomachean ethics, really throwing around terms, thinking like, could I be a classics major? And throwing around a lot of philosophy terms they're only half confident in. I think you're more confident than I am. I'm reading this, I'm like, yeah, I think that's right. I'm not 100% sure.

[B]

Sure.

[A]

But I liked it. I liked it. But it was a great piece that really, I think, frames the exercise of what they're trying to do in such an interesting light. So talk to us about what the Claude Constitution is and what it's trying to do that's most interesting in your mind, or what it does, what it says, and then what it may be trying to do beyond that.

[B]

Yeah. And so I'll make this brief because there is a lot of stuff you can get into and if folks are interested, and we have a law firm has another podcast series called Scaling Laws that I run with Kevin Fraser, and we talk about a lot of this stuff. We had a whole episode of Just on the Cloud Constitution that came out recently. So when you train one of these big models, the first thing you do is called pre training, where you're basically ingesting, feeding the model, the entire corpus of text in the world. So basically think of it as you give it the entire Internet, plus any book you can get your hands on. And you're asking it to basically develop a really complicated, sophisticated statistical model to predict the next word based on that. And that's useful, but that's only the starting point because you don't actually get something that useful out of that in the sense of something you want to talk to and can do things for you. You then have to go into what's called training or post training, where you take that giant text prediction machine and you try to get it to be useful. And again, simplifying greatly, essentially what you do is you come up with a big test suite of, hey, here's some user input, and here what would be a nice user output. And you come up with enough of those and then you get the model to sort of do that more and more. The pre training part is pretty similar across all the labs because you're ingesting the same corpus, because the corpus is everything. So there's not a ton of ways to differentiate yourself. The way you differentiate yourself there is basically by, you know, how much compute have you used. But on the second phase, this post training phase, that's where each Lab Google anthropic, OpenAI x meta, the Chinese open source labs, et cetera, et cetera, they really differentiate because you can get very, very, very different behavior based on the much more limited set of post training data that you give to the model. And so you end up with a lot of choices there on how you want the model to behave. Now all the labs have to think about this. All the labs make choices, and therefore all the labs at least implicitly have a vision of what kind of agent are they trying to create. The difference, I think with Anthropic is with Anthropic, I think they've done the most real thinking about what kind of vision they want. They've been the most transparent about it, they've been the most explicit about it, and they certainly have the most, I think, sophisticated and theoretical approach. Part of that is because who the personnel is at Anthropic, it's kind of sort of the philosopher nerd contingent of OpenAI that broke off. I mean, that's who basically formed Anthropic. And I obviously said philosopher nerd with immense admiration. And the person there who is chiefly responsible for not the technical aspects of the training of the model, but of this more theoretical is this woman, Amanda Askel, who literally has a PhD from NYU, which is one of the best philosophy departments in the world in moral philosophy. When you read these documents, and she was the main drafter of this constitution, when you read these documents, you're like, huh, this feels like a philosophy article because it's a philosophy article written by an actual reasonably professional philosopher. So that's one interesting thing about Anthropic and about the Constitution, how sophisticated they do their thinking. The second thing is, what does anthropic think makes for a good agent? Now again, sorry for turning this into a philosophy seminar, but this is kind of how you have to engage with this. When you think about what makes a person a sophisticated ethical decision maker. There are different ways of thinking about it. There are sort of rule based approaches or kind of calculation based approaches, and you have sort of Kantian rules is kind of one version of this utilitarian cost benefit calculation is another version of this. And these have traditionally been at least, you know, in kind of modern academic philosophy, the two main approaches. But there's a third approach that's much, much older. It goes formally speaking, all the way back to Aristotle, informally speaking, much, much, much further. I think sort of the wisdom tradition of the Old Testament and of Confucian ethics and all of that stuff is steeped in this. It's something called virtue ethics. And the idea there is that what makes someone a good and sophisticated ethical decision maker is not that they calculate or have these strict rules, but that they have certain psychological dispositions, to be honest, be kind, be helpful. And they have a kind of sensitivity to the facts to figure out how to do that, given that life is really messy. And what's interesting about anthropic and in this constitution, which again, I think is a bit of a misnomer, at least I think it's less of a constitution than what screenwriters will call a character bible, like a kind of psychological profile of this individual named Claude. What they're betting on on is that instead of trying to give Claude a lot of very specific rules beyond obvious stuff like, and they list this, don't create nuclear weapons, don't create child pornography, but there's not a very long list of those. Instead, we're trying to give Claude judgment, give it a list of interesting values to take into account. Though we're not going to prioritize which values. So if those values come into conflict, use your judgment. Their bet is that that's going to lead to the best agent. Which sounds crazy, except that when you ask people who spend a lot of time or kind of AI enthusiasts or the kind of Cognizanti or the Silicon Valley elite, which model do they want to spend their time with? It tends to be Claude, which actually often underperforms on sort of pure benchmarks on coding or math or whatever to let's say, OpenAI or Gemini. But people, and I include myself as someone who sort of lives in this ecosystem, spends a lot of time and just interacting with these models, the vibes again to sort of reuse that word from earlier the conversation. They're just better with Claude, right? They're just better with Claude. So to me it's kind of, I think it's really interesting about AI and the direction this is going. But in some sense, to me, what's most interesting about this is the sort of more academic, backward looking philosophical observation that after thousands of years of ethical debate between philosophers, we're almost running an experiment in silico in these new machine systems of what framework makes for the best ethical decision making. And be like, it turns out Aristotle was right. Or at least that's my takeaway from this. So again, there's a lot of interesting AI policy and stuff that we could talk about. But to me, as a former failed philosophy grad student, because I just was not smart enough for the discipline, so many people who got too scared of philosophy, they just went into law instead. I just find it so interesting that we're running an experiment and it's kind of cool. Turns out Aristotle was right after all. With all respect to Bentham and Kant.

[A]

I think it's really interesting. I enjoyed your piece because it kind of framed things this way. I have lots of thoughts, but I don't want to occupy the conversation. Molly, let me turn to you. I know you have spent some time looking at these models, writing about stuff. You did a lot of emerging tech work in your last world Washington Post. We will yet rope you into it here at lawfare, but we've been keeping you too busy with legal things and other random developments.

[B]

We're coming for you, Molly.

[A]

Yeah, beware, be wary.

[C]

I've been enjoying the bit of respite from the emerging tech world, but yeah, no, I find it really interesting too. I'm sorry if I'm turning down the sophistication meter here. I won't be invoking Rawls, much less Heidegger. I don't know who else you can bring into it, but I guess from the perspective of an everyday person who uses or thinks about AI, I found the Constitution and also Dario Amodei. I don't know how to pronounce his names. The anthropic CEOs essay on the subject of the Constitution that they've given Claude, but also generally on where these machines are going. He published this just a few days ago. I think it's called the Adolescence of Technology. I found it very interesting because I think a lot of people, when they think about AI, you think of the famous sort of hypothetical of if you told an AI okay, I want you to make as many like a robot, you want to make paperclips as efficiently as possible, and then it ends up killing everyone because that turns out to be the best way to make paperclips, and all it cares about are paperclips. Or if you think more recently, when Microsoft came out with its model Sydney, that said to the New York Times reporter, basically like, you should leave your wife and Love me. You know, there's all this sort of weird and kind of scary stuff that happens and, and you get to asking the question of, well, why is that happening? Did they just not tell it not to do those things? And so I think what's interesting here and what Daario Omodi acknowledges in this essay is that these machines are always going to develop some sort of Persona. And that might just be from kind of the raw data you gave them. Like maybe some of these were reading a bunch of science fiction where artificially intelligent machines turned against humans. So then they sort of start to talk about, about that stuff. But it's going to develop a Persona and then if you just give it rules, how it interprets those rules is going to be informed by that Persona. So what Anthropic seems to be saying here is, well, let's work on the Persona instead. And Alan mentioned a character bible. What Dario Modi says in his essay is it's like a letter to a child, like a pre sealed letter from a deceased parent. Like, read this after I'm dead. This is how I want you to be as a person. And so kind of weird, kind of spooky, this is my child vibe to that. That's a little off putting. But. And of course that also leads to the question of but whose ethics are you giving it? If he thinks he's the parent, is it his ethics? And do we want this one person or these few people who are developing these models ethics to be substituted in for all of our ethics as society, which we can't even really agree on, which Alan gets into a little on his piece. But anyway, I think it's a very interesting, very different approach and it's designed to avoid some of the traps that happen when you don't go to the effort of telling these machines what their Persona should be when you give them rules and then just leave the Persona to develop on its own? So he gives this one example where they tell Claude, don't cheat, don't reward, hack your training environments. But it was trained in environments where hacks were possible and it had done some of the hacks. And then Claude said, I must be a bad person, and then started doing a ton of other destructive stuff, right? So this says Claude, you're not a bad person, you're a good person who has to be balanced about things to adapt to the realities and context, blah, blah, blah. There's no way of knowing exactly how it's going to adapt. You know, it's still going to develop a Persona. That we can't completely control. But going at it from the perspective, from the point of Persona rather than the rules that will then be interpreted by said Persona, I think is kind of fascinating.

[A]

It's a fascinating enterprise. And tell me if you all think this is wrong, but this is what jumped out on me and think about that. I will say, having skimmed the Constitution, read some secondary stuff, I want to spend more time on it when I have not two stir crazy snow bound children terrorizing my house. So if I call my Aristotle correctly, a big part of Aristotle, I think I only ever read the Nicomachean ethics with any particular close look. But a big part of it is about essentially conditioning. You're about building good habits. Virtues is about embedding good behavior. So to some extent what you're trying to cultivate isn't this. It is whatever the product of this is. That's why that metaphor of the letter from a grandparent or a parent, I can't remember exactly what he said in that piece actually makes a lot of sense to me. Me, because you're saying like, oh, this is something about informing my perspective in engaging on something. But really it is just one input, perhaps a fairly foundational input. I think that's what you say constitution that you're implying, right? But it's really the outputs, the downstream conditions that come from how that information gets translated and conditioned and built into my interaction with the outside world that establishes actually my virtues. Right. Virtue isn't established. I don't take to be something you can define even in vague terms on the outside. It's actually something you encounter through practice and applied reason and conditioning. And that actually makes a lot of sense for me here. Right? Like you have a rules based system that only works so far as you can't. You have an entity that's not trying to out reason the rules and look for technicalities or look for ways to break the rules or manipulate the rules.

[B]

Right.

[A]

We've learned this today in our rule of law in this country. Right. The, the real question then becomes like, if this is really what you're looking for, what you're trying to control is not what this directly controls, but downstream outputs. It has to be part of a process. Right. It's less about what just this constitution is, but now about how Claude will be evaluated, educated, conditioned and interact and evaluated moving down the line. It has to be on ongoing enterprise. I think because this is just, just one input it's going into. It's not actually a sign of there being a virtue because even with this you don't know how you're going to apply it. So I guess it's like the US Constitution, like any constitution you have a constitution, but then there's an er constitution like a proto constitution, the most important part of it. And that's the rules for mending the constitution. Right. That's the part where foundationally actually determines the content. And that's why our constitution has changed once it was initially enacted here. That's the part that's less clear to, to me. When do they come in and have correctives or talk to Claude or do that other socializing conditioning that comes later. They may not have a clear strategy for that. Maybe that's reckless to have a clear strategy for that. But this strikes me as just one part of a much bigger process that's actually going to find what they're output on and it's like a guiding light and maybe a metric, you measure it against as best you can define it. Am I crazy about this? I just feel like this is a part of the process but really what we're talking about is much still like a multi year thing that this is actually a substantial starting point but kind of a bit player in the other variables that we just don't know yet.

[B]

Oh yeah, absolutely. And so the question is how does this get integrated into Claude 5 and 6 and 7 and 8 and then at some point once these models no longer have to be explicitly retrained for the next generation but once these models develop what's sometimes called continual learning, which is the ability to go out into the the world, whatever that means, but through interactions with users and once they're embodied through actions with the physical world to be able to sort of tweak their own weights and biases and parameters. And so yeah, I'm not sure to be clear, I don't think Anthropic knows. Right. I think one things I like about Anthropic, and again I suspect this is partially from the fact that Amanda Askel has this academic training and philosophy which is very much a discipline and which not over claiming and being very clear about what you do and don't know is a good thing. And I suspect that people I think around her have a bit of that same culture as well. One thing I appreciate about Anthropic is they're quite sort of transparent in the hey look, we're doing the best we can. We're trying to build a machine. God, we have no idea if this will go well, but if we don't do it someone else will. And at least we're trying to do this properly. I think they don't know the answer. But yes, to your point, Scott, no one does not just do this one time. Right. This is both making sure that Claude currently is behaving as well as it can, but also to guide Claude and anthropic more generally as you go forward. How do you build a more and more and more sophisticated discriminating faculty of mental judgment, of moral judgment?

[C]

Yeah, I agree. I don't think they know exactly what they're going to do. What they've said basically is that you combine the constitutional AI with better interpretability, that they're trying to improve their ability to look inside these models, which everyone says are black boxes, but that's not entirely true. You can kind of map their reasoning circuits to some extent, et cetera. And then you figure out, okay, this is what we wanted Claude to be when we wrote this Constitution. Is it acting that way? And then maybe a little bit you can get into why. And then they've said, then we'll go back and we might revise the Constitution and then we try again.

[A]

Yeah. The other part of this that really jumped out at me that I thought was really interesting is where it talks about Claude's own moral status. This and the bolded subject header of this is, Claude's moral status is deeply uncertain. It's really, really an interesting idea. Something I think is a little bit people have resisted talking about except in deep AI circles. People dealing with AI more from a policy perspective, been like, that's such a over the horizon hypothetical. This question of AI itself being an intelligence of moral status, potentially at least, or at one point perhaps being there. And it's interesting in that it doesn't say it is necessarily or will be, but it definitely leaves the door open as a possibility, acknowledges that as a possibility. I'm curious about how you all respond to it. I find that, personally, I think I said on the podcast before, it's been a couple years. I find that very important in part because I'm an almost lifelong vegetarian at this point, who thinks moral failings come from. We fail to evaluate the moral value of other creatures that just may not be like ours. And there's an ethical, ethical reason. It's part of the reason I'm a vegetarian. I read Animal Liberation way too seriously when I was 17 or whatever.

[B]

Peter Singer got you.

[A]

He got me. He got me. And, you know, a lot happens. A lot of people read that book, I gotta say. But I do think there's something there to that, which is like, if we blind ourselves to the moral capacity of these actors, you really run the risk of doing something horrible. And a couple years ago we had those stories about the Google engineer who was like, I really think this AI is intelligent and we need to start treating it like an intelligence. And everybody kind of laughed at the guy, kind of laughed him out of the room. But here that's actually being taken seriously by anthropic, which is, I think, refreshing, at least that acknowledgment. But I'm curious, Alan and Molly, how does that come into play and what does it mean for this whole enterprise? Because can you really pretend that you can? Part of being a parent, right. And training a child is acknowledging the fact that you may fail and that child may end up horrible and they have a right to be that way. So how much does toying with this idea of there being moral status of this intelligence limit our ability ultimately, perhaps in how exactly we can shape it or limit it or decommission it at some point?

[B]

Yeah, I mean, look, on that last point, I'm not sure that children have the moral right to be horrible. They have the moral right to be themselves. But it's an interesting question. Do they have the moral right to be horrible?

[A]

By my extremely high standard, they can depart from the lessons I give them and I can't enforce that on them for the rest of their lives.

[B]

But I think that's an interesting point.

[A]

There are limits on that.

[B]

Look, I think you're right and I think you bringing it for vegetarianism is actually a good point. And I actually would expect that vegetarians and vegans actually probably will take this a little more seriously, this possibility. You're right, that policy people often don't talk about this. I actually don't think it's because it's that speculative and so far away at the rate that these models are advancing. I don't think it's that far away. It's potentially a matter of years before we have to take these seriously. I think the real reason people don't talk about this is actually because of the last point you made. Made because once you start talking about it, it's hard to talk about anything else. And it's hard to. It swamps all the other questions. Right. Because if it's possible that at some point these models will be of moral concern, and I think that that possibility is not just possible, but almost obvious. And let's go back to the animal point for a second. Right? You don't have to be a vegetarian or a vegan. Vegan to think that animals have some moral concern. If you have a dog or a cat as a pet, you already almost certainly believe that. And even if you're a meat eater, right, as I am, but spend a few extra dollars per egg carton, right? Or, you know, buy the not horrifying, you know, industrial chicken product in the store. Because even though you have decided or have convinced yourself that eating chicken is ethical, you know, maybe the worst, you know, the worst extremes of chicken cruelty are not so good. I think you've committed yourself to the proposition that sentience, or at least whatever counts as sentience for moral concern is not a binary. And like, if a chicken can have moral concern, why can't an AI system that is already and certainly in the future a lot smarter than you are on some and soon to be almost every intellectual domain, why can't that be of moral concern? Right? It's not like, and frankly, if you're like, well, they can't be sentient because they're not, I don't know, carbon based. They're, they're not meat based. Right. They're not gooey between the years in the way that we are. I mean, that's a. You're, you're, you're. There are a lot of smuggled assumptions that we understand how consciousness in human beings works. So I think Anthropic's position is the only intellectually defensible one, which is, look, we have no idea, but we certainly can't rule it out. Now let's imagine in the future that we decide that these things are sentient either because we think they really are or because they're so behaviorally sophisticated that we can't help as people, ourselves psychologically that like to project sentience onto other entities. We kind of, you know, think that our AI friends are sentient in some way. What do we do about that? That's actually not entirely obvious because, you know, it's not clear that the result is we have to give them total, complete autonomous economy. We have to treat them well. But of course we're in the odds position of getting to write their utility functions. So, you know, just as, you know, if you. And again, this is not a perfect analogy because dogs are less intelligent than humans are, though in some ways they are more intelligent, you know, and like smelling and stuff like that. You know, if you get a working dog, you know, on the one hand you might say, well, it's unfair that I make the dog work, but the Whole point of working dogs is that they like to work. Like that's what they like to do. So one can imagine a world in which we have these AI entities that we have designed to like to help us. Now, was it unfair that we designed them that way? Do we have to give them the ability to be fully free? It gets very complicated, but I think there are possibilities between. We have to give them total full autonomy, and if we've done anything else, we have created sort of a slave race of incredibly intelligent machines and we're all going to hell for it. But I don't think you can, like, not think about that issue, is what I would say.

[A]

That is refreshing.

[C]

Yeah, I agree broadly with that. As far as the question of whether it has moral status, I guess I think kind of two separate things. One, the point at which we don't have control over it would definitely be some meaningful turning point there. Right. Like you can give your children, and we're going to the child parent thing again, whatever values you want to give them, but eventually they're going to develop on their own, independently. And then two, even so, it's a little different from a child because we didn't actually design our children. Right. We didn't actually, like, code them in any way.

[B]

Speak for yourself. I have a very elaborate Skinner box that my kids are very. They delight, they love it. So it's fine.

[C]

I don't have kids, so I'm actually not qualified to. I have no idea. I'll find out eventually, maybe, you know, whether. Whether I can code them. It sounds convenient, to be fair. So anyway, I mean, I think it's interesting. I do think that the point at which we acknowledge that these have moral status, I would imagine is also the point at which we're acknowledging that we cannot control them, at least entirely. Which, you know, we may already kind of be at that point, but. But we acknowledge that they've developed into something that's independent of us. And then I don't know where that leaves us. Right.

[A]

Fair enough. Hey, folks, Scott R. Andersen, cutting in here from the future with quick note. We ended up spending so much time talking about our first two topics that we decided to hold our third topic. It's going to be something about TikTok for another time. So just two topics for today's episode. Thanks. Well, folks, that is all the time we have. But this would not be rational security if we did not leave with some object lessons to ponder over in the week to come. Eric, what did you bring for us today?

[D]

There are people in my family who experience migraines. And one, I brought something that has on occasion helped my wife, which I found kind of on Amazon many years ago and is called the Tingler. And you can. There are many varieties of this. You can find them on Amazon.

[B]

It's actually my, it's my professional wrestling stage name, just so you know.

[D]

Well, exactly. So there's a kind of a nice kind of tie in. And if you use the affiliate link, then Alan gets 10% and you, you put it on your head and it just kind of rubs it. And for some people it's. Migraines are a very kind of complex thing and everyone's experience is different. But for some people, sometimes this, this helps. And I should say that I was asking my wife earlier today, what, what object? I said, look, I'm going on the object guys podcast. What object should I find?

[B]

And.

[D]

And she suggested this. So there you have it.

[B]

So first of all, two things I just want to say. First, everyone should watch the YouTube video of Eric self tingling. And then second, I just love. I. I don't remember the last time someone actually brought an object for their object lesson.

[A]

It is a very literal interpretation of object lesson and I appreciate that. I do think that was the original. That was the original model 10 years ago when this podcast started before I was on it. We've kind of drifted from that in the ensuing years. But that sounds great.

[B]

I just do TV show.

[A]

Sounds like Molly. There you go. Maybe I have another physical object coming up as well. Well, Alan, what did you have for us this week if not a physical object?

[B]

Yeah, I'm so lame. I don't have a physical object at all. No, I have a TV show show. My wife and I have been watching season two of the Night Manager with Tom Hiddleston who is just. God, he's so dreamy like it is. It's a problem. I mean my, the, the man. The man can wear a suit like nobody, nobody else. It's great. It's really fun. I think I'd sort of watch Tom Hiddleston in anything. He's just so fantastic. But if, like, you want a good, stylish Tom Hiddleston esque spy thriller, strong recommendation.

[A]

I think this is the second week in a row we've got a Night Manager endorsement.

[C]

Oh no.

[A]

Did I repeat behind that last week? I actually didn't hear it because Anna recorded her separately because she had to leave a little early and sent it in. But I think for the show notes it looked like that was her recommendation. But if I'm misremembering. Forgive me. I haven't gone back to listen to the object lessons at the end of last week's episode since we recorded it. Well, for my object lesson this week I'm going to come to you with two winter related endorsements because I have, as I mentioned several times, been snowed in with both my children and I'm going slightly insane and can't really think of anything that's not snow or ice related at this point. One is something called the Fire Snake. I still don't know what exactly it is, but it appears to be the wonderful people at Metro have been using it to thaw ice so that they can repair rail lines. There's an amazing, I can't remember on Twitter or Blue sky one of the two threads out there showing video about how they essentially light a whole rail of track on fire to melt all the ice around it and then can fix gaps in the track. It's pretty amazing. Somebody, I can't remember who for the life of me at this point described it as casting a level three spell. So why the hell can't we get the plows to work? Which I think is a great, great way to think about it because it really is quite extraordinary.

[B]

This is your second D&D reference in one episode, by the way. You hit three and you don't want to know what happens when you hit three.

[A]

I think. I think it's a crit.

[B]

Let's get ejected.

[A]

I think we're there. But regardless, it is a really, really fascinating video. I highly recommend checking out a good tribute to the wonderful people at Metro. I love the metro around D.C. even though I don't use it. I could do anymore. My neighborhood doesn't have great access to it, but they do really hard work, especially this time of year. And it's good to see them getting some credit and much appreciation because they've been doing pretty well during this winter storm compared to a lot of other public transportation. My other one I'm going to endorse now know it's long underwear, guys. It's great. It's wonderful. I bought a very expensive pair of smartwall long underwear last winter. Two winners ago. And I gotta say I've never regretted. I think I've worn it like maybe half a dozen times and I'm wearing it now.

[B]

I'm wearing it now. I, I took, I took the dog for a walk right before Rat Sack and so I threw, threw on my long johns and I'm toasty, man. I'm Feeling great.

[A]

I had to take it off because I only bought one pair and after four days of shoveling and needed a little bit of a break in a wash. But I'll be back in it as soon as I get it out of the dryer. Highly recommend. Well worth. Worth the investment. For anybody in the mid Atlantic and below who like me, may not have thought that was a natural purchase to have and it seemed ridiculous to spend that much money on. I don't regret it strongly endorse. Especially if you bike back and forth to the work. I will say that's when I use it more than anything. And with that, Molly Roberts, please bring us home. What do you have for us this week?

[C]

Yes, I've brought a hat. Not a cat, although he's still there. So I was thinking about this mostly not in the context of this conversation, but in the context of a conversation I had with, with Ari Tabatabe, who's a lawfare public service fellow about NATO and the international order and kind of America's. The United States's abdication of its role at the helm of that. So I have this hat. It says I'll put it on maybe. Although will I be able to read it on the screen? I believe it says, hello, I am fleeing the century of American humiliation. Can you show me where to buy white monster energy drinks? And then it has it translated, I think, into Mandarin underneath that. So it's hat to wear when you flee to China, which I would not be doing anyway. Don't ask why I have the hat, but I have the hat. And I was kind of thinking, thinking about that.

[A]

That's a great hat. I don't know what white monster energy drink is. I'm not sure I want to. As somebody who some might call a white monster himself, I don't really want to know.

[B]

Scott, you're way too old. We're way too old for beverages like that. Our little hearts would explode. We're not. We're not young men anymore. We can't be drinking shit like that.

[C]

They have versions of the hat for like, can you show me where to buy Flamin Hot Cheetos as well? I don't know if that's past your.

[A]

Time or that is a little more up my alley. Well, regardless, a wonderful hat.

[B]

Direct me to the soup dumplings. I would wear. I would wear that hat if that's what it said at the end.

[A]

You're just pandering at that point. Alan, come now.

[C]

Right. No, I mean, this is maintaining the American spirit even, even in the Century of Humiliation.

[A]

There you go. Well, with that, that brings us to the end of this week's episode. Episode Rational Security is of course a production of Lawfare, so be sure to Visit [e-mail] for our show page, for links to past episodes, for show notes for our written work and the written work of other Lawfare contributors, and for information on Lawfare's other phenomenal podcast series. While you're at it, be sure to follow Lawfair on social media wherever you socialize your media. Be sure to leave a rating or review wherever you might be listening, and sign up to become a material supporter of Lawfare on Patreon for an ad free version of this podcast, among other special benefits. For more information, visit lawfaremedia.org support our audio engineer and producer this week was Noam Osband of Goat Rode Under. Music, as always, was performed by Sophia Yan, and we are once again edited by the wonderful Jen Patcha. On behalf of my guests Eric, Molly and Alan, I am Scott R. Andersen. We will talk to you next week. Until then, goodbye. We interrupt this program to bring you.

[C]

An important Wayfair message.

[B]

Wayfair's Got Style Tips for Every Home.

[A]

This is Stiles Mackenzie helping you make those rooms sing. Today's Style Tip when it comes to.

[B]

Making a statement, treat bold patterns like neutrals.

[C]

Go wild like an untamed animal. Print area rug under a rustic farmhouse table.

[A]

From wayfair.com fierce this has been your.

[B]

Wayfair style to keep those interiors superior. Wayfair Every style Every home.