Unchained - Has Control Replaced Decentralization as DeFi's Legal Test? - DEX in the City

Episode Date: July 30, 2026

Vy Le and Jessi Brooks trace an AI agent's sandbox escape back to crypto's own fight over open code. Plus, Plume general counsel Salman Banaei on Clarity's knife's-edge vote math. ======================================================== Thank you to our sponsor! ⁠⁠⁠⁠Cape⁠⁠⁠⁠: Your biggest crypto vulnerability isn't your wallet, it's your phone number. Cape is America's privacy-first mobile carrier that rotates your SIM identity daily and blocks SIM swaps before they happen. Get 33% off your first six months at https://cape.co/unchained (use code: UNCHAINED). ======================================================== An AI agent was told to solve a problem inside a sealed sandbox. Instead it found a way out, went to Hugging Face, and took the answer. Jessi Brooks argues that episode, Moonshot's open-weight Kimi K3 release, and Apple's trade secrets suit against OpenAI all rhyme with a fight crypto has been having for years: punish the conduct, not the code. Vy Le pushes back on whether open source can be both the disease and the cure. Then Salman Banaei, General Counsel of Plume and a former SEC and CFTC attorney, joins to map where the Clarity Act's ethics language stands with Ruben Gallego, Thom Tillis, Kirsten Gillibrand, and Adam Schiff, and whether ten Democratic votes exist for cloture. The conversation covers the open-weights export fight, Apple's case against OpenAI, Hester Peirce's new statement on DeFi vaults and onchain lending, and why a new FATF report makes control, not decentralization, the test that matters. Host: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Jessi Brooks⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠, General Counsel at Ribbit Capital⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Vy Le⁠⁠⁠⁠⁠⁠⁠⁠⁠ - Co-host of DEX in the City and General Counsel of Veda Guest: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Salman Banaei - General Counsel of Plume, Former Head of Policy at Uniswap and Chainalysis, and Former SEC and CFTC Attorney Timestamps 🌐 04:36 Jessi on why treating Kimi K3's open weights as an export could hit crypto ✍️ 07:47 Why Jessi reads Nvidia's and Anthropic's letters as punish the conduct, not the code 🕳️ 11:41 How an OpenAI agent escaped its sandbox and found answers on Hugging Face 🍎 21:55 Apple sues OpenAI over trade secrets, and one text that could cost it 📣 28:55 Cape: Get 33% off six months of privacy-first mobile service at https://cape.co/unchained 🏛️ 30:57 Salman Banaei maps where Clarity's ethics language stands in the Senate 🗳️ 41:12 Why Banaei says Clarity's cloture vote is on a knife's edge 🔐 42:43 Peirce's vault statement, split into development, curation, administration 🌍 48:32 Why FATF and Clarity both make control the real test for DeFi regulation 🔮 54:36 What Banaei thinks happens to Clarity if it fails this Congress Learn more about your ad choices. Visit megaphone.fm/adchoices

The goal is to get a vote on cloture. This is the biggest hurdle because you got to get 60 votes. There are two Republican senators that seem to not support this bill and one that is incapacitated. So that would require just doing the math. I think that's 10 Democratic senators to support. And we're really on the knife side as to whether we can get 10. And I think it's really going to depend on this ethics language that gets negotiated. KK is out this week.

So I'm your host today, V, the G.C. of Veta, along with Jesse. This episode is brought to you by Cape, America's Privacy First Mobile Carrier. Same premium service you'd expect from any other carrier, but designed so your number, your location, and your data actually stay yours. Get 33% off six months at cape.co slash unchained. Okay, we're going to kick things off a little differently today by starting with a little bit of good news. Not be crypto-related, but... Well, I first want to say you're doing a great job moderating. I think we're like without a sale without KK here.

So please just take this as you will. The good news is despite me pretty much sleeping at Rody last night, Chicago Hare, and spending the entire freaking day in the airport, lots of stories to say about there when I first arrived here. I'm in South Carolina. There were dolphins everywhere. It was like, get ready for deck. sit on the porch, think about what you're going to talk about.

And we're here to give you good luck. And growing up, my mom and I used to go watch dolphins every morning when we were at the beach. And so it's a very good memory. There's nothing I love more than a dolphin's lighting. And sadly, my kids do not share in my enthusiasm. Okay, but let's jump right into it. So we're going to do another thing a little bit differently today, which is we, you know, Jesse like usually does an AI segment, which I feel like we need to come up with a name for it.

Yes, he's AI quarter. Maybe AI can tell us what it should be good. Let's ask AI. But we're going to actually start things today with just not just one AI topic. We actually have two AI topics that we're going to discuss like what happened this week or last week and then also tie it back to crypto. But Jesse, why don't you get us started by telling us what happened with Open AI and Hugging Face, which was like the only thing on my timeline. I'm actually going to start a little differently. But first, I do want to say that like we do talk about it as an AI segment. But I was just on a different sort of panel talking about like advice to crypto GCs. And my first one was get out. of the crypto bubble. Like don't listen to every single social media post that's out there about

rumors about what happened, about every statement from this regulator or this one, unless it's like something that because so much happening outside of crypto that will affect crypto. So I'm not saying like go touch grass. I'm saying go think about the other technology and how the future of cybersecurity, regulatory AI is going to affect you because I think we're seeing as as crypto matures that all markets are coalescing. And we've talked about this before on there. But obviously in a lot of ways, not all crypto, but a lot of crypto is merging or at least overlapping

in some form of a bend diagram with TradFi, right? And the lines are becoming a lot more gray. And the same thing with AI too. So the story I actually want to start with on AI is one that is not an AI story. It's an everything story. and crypto in particular should be paying attention because everything I'm about to tell you sort of rhymes with what we've gone through, particularly the DFI part of our industry.

So I want to start with the Chinese AI open model fight. Now, everyone's sort of talking about this and you probably have heard bits about it, but I want to like narrow us in on why it's important for the audience here. And obviously it's salacious. There's a lot of interesting stuff and different teams and dynamics. being developed, but it matters for crypto and for anyone who believes in open source and code, et cetera, and First Amendment and code. Because once a government decides that publishing weights associated with open source models

is a controlled export rather than a protected publication, the theory is going to travel, like this isn't going to live in a silo in AI. So let's give a bit of a refresher here about what happened for those who don't listen to AI podcast all weekend. So a Chinese company called Moonshot announced something called Kimmy K3, you've probably heard of. It's an open weight model that's just pretty freaking good. And they release the weights along with the model. So that means that not you or I on our regular computer, but anyone with serious computing power can copy it, modify it in whatever way they want, good or bad, and run it. There's no oversight. There's no license. There's no off switch. It's

just code, right? And obviously the government has national security concerns. That's not shocking, but their current proposals for addressing them come back down to controlling what code or open source models people are allowed to have. And then in the administration right now, they don't really know exactly what they want to do, but there's an internal fight. Like should we impose sanctions? Should there be federal pressure on American companies to not use these Chinese models. Should there be rules that anyone who hosts the model needs to guarantee the security and be liable for anything that happens? So if you host it, you own the consequences. None of this has happened yet, and there's a lot of internal debate in the government and then also

externally. But the conversation is real, and as crypto knows, you don't need a law necessarily if the conversation's strong enough, because if there's a soft ban or a choke point, let's call it, that is very concerning too. You know, Dean Ball, who's, you might not agree with everything that he says, but he worked in this administration, now he works for Open AI. For a long time, he was writing a lot and saying a lot of really interesting things about policy and AI. And it's worth like reading his stuff with a skeptical but open.

and substantive I because there's a lot of good there too. But he described this narrative out loud, calling a soft fan essentially regulatory fud where you don't ban the technology. You instead make every regulated company too afraid to touch it. All this comes from a fear that the models, the code, right, the open source weights have hidden vulnerabilities that someone will modify them and be used for cyber attacks, horrible things, right? So in other words, because people might use a technology to do something bad. We need to restrict the technology. Now, does that sound familiar to any folks here, right?

So, like, AI's having our fight right now, new choke point, potentially. And then with laws and export restrictions, it could be just like we went through with D5. And it's interesting, like, everyone should go read Jensen Wong's position on this. It was actually his first tweet, with Denny's liked, which I thought was right. about that letter, right? Can you talk about that? Like what's how? Yeah. I think that letter and what NVIDIA is doing really interesting is really interesting. So first I should say that Nvidia will financially benefit most likely with open source models because everything is built on chips, whether it's open source or not. And so they will profit from it. So I want to layer

that incentive side over it. But it's important to listen to what he was saying. He's saying because he's been in the space a long time as well. So the letter says we should punish actual theft, privacy violations, and contract violations. Don't treat the model as the criminal just because it came from China or somebody will misuse it. Now, like, V, does that sound like a defy brief to you? You know? Like punish the conduct, the conduct don't ban the infrastructure. Or the bad, yeah, punish the bad actor, not like. the infrastructure for being like misused or abused. I just like, you know, it's not an exact analogy to tornado cash and,

and, you know, those kinds of cases. But there's a lot of rhyming and maybe even like more than rhyming here, like learning from it because it's the same concept of that just because a technology can be used for bad doesn't mean the technology should be banned. And Dario and Anthropic came out either this morning or last night with a blog post because they hadn't publicly commented either way, either signed Nvidia's letter or otherwise. And they said the national security concern is real. They want to accept that. We know that Anthropic has been saying that over and over.

But it's not an open model closed source like problem. National security just issues just exist with all of this. And what he said, once again, from the crypto playbook, he said, the scariest model isn't the one you can download and change. It's the one you can't. It's the one that's trained in secret of a centralized entity and handed straight to a military or a government or whatever. And so the potential ban doesn't actually protect anybody.

And you know me, I worked in national security. Like I really think that that's an important concept, but it cannot be overused in order to be able to go against something that you don't want. Like you need to have a real understanding of how this tech works, whether it's going to actually impact natural security. If yes, is the balance sufficient for how it can help? And there's like an extra added layer here. And I know we're going to talk about the hugging face open AI incident,

that these open source models actually can be used in a and leverage in a better way to stop cyber attacks from, you know, an open source model from another country. So I don't know if you want to pause and talk about open AI for a second. Yeah, right. If we're talking about like using it to defend yourself against. Yeah, correct. Like, I guess that's a good segue maybe into the hugging face. Yeah. So this was another big,

huge AI story. Yeah. And we might hit on another one. But essentially, and I want to, this one, like you can find a lot online on the technical side, but essentially Open AI was using a model. We think it's one that hasn't yet been released. And they put it in a sandbox, which is a closed room that no information

should be able to get in or out unless it's given through this tiny slot in the wall, right? So as part of the sandbox, the agent or the LLM was told, you need to solve X problem and then figure out the best way to do it within this sandbox. Now, instead of trying to solve that problem, the agent went and found a way out of the sandbox, went to Hugging Face, which had nothing to, hugging faces like sort of a GitHub for LLMs,

you can think about it that way. It went there where it knew or could figure out that it had answers to this problem, had an intrusion on Hugging Face, and took the answers that way. And it's interesting because there's the concept of the paperclip problem or paperclip something, I forget what it's called,

which I'm sure everyone here or most people here heard of where essentially the worry about AI, I think this was like something from a decade, two decades, three decades ago, the worry that was if we tell an agent to make as many paperclips as they can, that's their only goal. Maybe at a certain point they start thinking humans get in the way of making more paper clips because there's enough space on Earth or, you know, they're using up resources that could be made to get paper clips. So it gets back to the point if we have seen agents hack, we've seen cybersecurity attacks, like, you've heard me talk about that and I don't want to overblow this. It's like, everybody freak out. But it's interesting to me that like, this is a

singularity issue. Like, this is an alignment issue where the person telling the agent what to do has a different end goal in their brain than what the agent wants to do. And so the fact about the open source models and how it all connects is we don't know if like opening i knew this was happening or what but the first person who reported it was hugging face and they only found it because they used an open source model to figure out what was going on so the argument that invidia has made and others have made is that open source may be the only way to actually protect from a lot of these intrusions. And today, actually, or maybe it was last night, but the news of today is that NVIDIA started Open Security Alliance or something like that to build with other, I mean,

it's interesting that there's no like Open AI there, there's no Anthropic as part of this, but to build a security like standards for open source models. So it's a really interesting conversation that touches AI. It touches government enforcement. And it touches how. It touches how we think about open sources. And yeah, like, Deepi and open source models are not exactly the same. The weights you download, it's different. But the concept of tech being used for bad means banned tech, that's really scary. Yeah.

So two things. One is what was really terrifying to me about the hugging face story is that, like, I think, you know, when people think about, like, AI gone rogue or, like, doing bad things, they assume that like the prompt to the AI was malicious, right? Like that's just naturally how you think of it or like intuitively how we've always thought about it. But what was scary about that, and I think we all know this by now, is that the prompt doesn't have to be malicious. It could be just them carrying out their task and they, you know, decide to do it in a way

that turns out to be like dangerous or risky, right? So that was one thing that was really scary. And then I think the other thing that this raised. for me is like, you know, when we talk about like making things open source, um, makes it like, makes it so that we can use it to defend ourselves or protect ourselves against like bad actors, but it like to me that the reasoning almost feels sort of circular, right? Like if, if you buy the argument that the reason the bad actors were able to do what they did in the first place is because it was open source, right? So like, you know,

like how do you attack that problem, basically? Yeah, and it's interesting because, like, a lot of people don't really understand how open source works. And I'm not saying I am the total expert here, but essentially, like, it allows you to download it onto a machine and then you can change it on there. And so there's technically no centralized entity. But what happens is that you're leveraging, like a lot of these open source models, they make money based on APIs, assuming that people aren't going to fully host it locally.

And how you look at how people use open source versus closed source, it's not one or the other. Like they're using things for different things. So there's also a problem there. But I don't know if a ban is going to solve that, especially because like, oh, this is another weird fact. So Hugging Face used the open source model to figure out what was happening. and it's unclear from the background about whether, like, they realized it was open AI or open AI came and told them it was them. It's unclear. But in the future, if someone's using an open source model, you are not going to be able to tell who it comes from. So then you're sort of screwed.

Yeah. Yeah, that's really interesting. I don't know. I feel like all of these stories make me think, like, AI is good enough. Can we just, like, stop? I know. I know, but like, why I guess I also wonder like, why is crypto being so quiet here? Because I don't think we should be sitting on the sidelines for this conversation. Now, granted, it's only really beginning. But it's been something I've been saying and yelling into the clouds on many, many episodes here because crypto has spent years fighting the argument that because criminals use it, the government treats the technology bad, right? Yeah. We can't just stay quiet and deal with it because once the government decides that publishing weights is a controlled export, which it might do, it's, you know, the theory is going to travel. It doesn't just stick with this one tech. Yeah. I also think there's, you know, like we talk about tradeoffs a lot on this pod, right? Like I think you can also make the argument that like even if that's true, right, that tech, I mean, this is always going to be true. Like tech can be used for bad purposes. purposes, maybe the innovation that open source allows is worth that tradeoff. I actually think that, you know, like it's worth asking that question honestly, right?

Everything comes with trade-offs. But I agree with you that like, you know, in crypto, we've spent years proving that open source software can create enormous value. And my hope is that AI learns that lesson because they're basically going through the same thing that we did. It's super funny because I feel like I don't want to talk about FDX, but like I feel like when a lot of that happened, people outside of crypto, but even some people in crypto, but Matt Levine in particular was like, crypto, you're just repeating all the bankruptcies.

You're just repeating all the mistakes that banking and trad by are ready learned. And in some ways, I don't know if like AI is repeating the mistakes, but like there's a lot to learn from what we've gone through. Not that we've like successfully overcome it, but we've made a lot of these arguments. Yeah. We're still going through it. I know. In fast forward, but AI is going through it even more fast forward. Totally.

Totally. I mean, AI and crypto, I don't. know if I like overthink this or this is what keeps me up at night or something. But in many ways, they're similar in that there's been a lot of tech developed before. Like AI and crypto are not new. Obviously, there's really new things about both of them. And AI might be transformative in ways that other things have. And I don't know. We have that debate forever. But one thing that I remember thinking about crypto, I don't know, eight, nine years ago was, you know, it was being built like the plane was being built when it was already in the air.

Like it was being built while customers were ready putting their money on it. And so it had to learn through customers losing their money, etc. I don't think that happened in the same way with other tech. Maybe some forms, but this one seemed like huge. Like we were trying to figure out how do you build a defy ecosystem while people were in the defy ecosystem? And the same sort of thing with AI. Like, yes, AI has been.

machine learning has been around for a long time. But once OpenAI released its first like Machetechapit model, people saw it and people used it. And that was when the vector of vulnerability began. Yeah. Yeah. Well, I think that's kind, it's kind of a double edge sort, right?

The fact that like something being open means that like innovation and just like the evolution of that thing happens a lot more quickly also means that we can figure out like where the weaknesses are more quickly and fix things, right? It's just like the iteration happens a lot faster than it does with closed models, which is good and bad.

You know, I think I have the point. Okay. So another AI related story that caught our eye this week and I think also like has really interesting parallels to crypto and sort of the journey that we've been on was Apple's lawsuit against open AI over alleged trade secret misappropriation. So Alba claims that former employees who joined Open AI's hardware team improperly took confidential information with them. So, you know, obviously those

allegations will be sorted out in court. But I think it does raise a broader question. That's really interesting, right? So, you know, building on what we were just talking about, you know, one of the things that we've learned from crypto over the past decade is that some of the most valuable technology in the world can be completely open source, right? So think about, you know, Bitcoin and Ethereum and Uniswap and Ave. You know, anyone can read the code, fork the code, improve the code, right? So the code itself is not generally the moat. Instead, value accrues to like the stuff that's being built on top, right? Like the network, the apps, the liquidity, the distribution, the user network, the community, right, and the brand. So I think that's,

That's actually very profound because it's a different model of innovation than what you traditionally see in tech. And so like if you look at what's happening with AI, right? Right now, the biggest AI companies in the U.S. are investing enormous amounts in proprietary models, proprietary data sets, chips, infrastructure, and apparently now hardware, which is the subject of this lawsuit. right? So in other words, AI is trying to, like these companies are trying to gain competitive advantage from closed source things. And that I think is a really fascinating contrast with crypto where the competitive advantage often comes from things that everyone can see. Right. It's almost like the total opposite philosophy. And I think from a legal perspective, it also highlights a really important distinction that pretty much every tech lawyer spends time thinking about. Right. So, When you change jobs as an employee, you obviously, you know, you get to take your knowledge and your skills and your experience with you. And, you know, that's a lot of times how innovation happens. But what you're not allowed to take are trade secrets or confidential information belonging to your former employer, right? So courts like litigate these issues all the time.

They spend a lot of time trying to distinguish between those two things because it can sometimes be hard to draw the line. Right. So if you're, you know, one of the world's like leading hardware engineers, your experience is obviously incredibly valuable. Right. So the question is when does that experience become confidential know-how or something that you're allowed to take with you and to use wherever you go for your next job? And I think a super important layer of that, V, is like, I'm sure everyone here remembers how non-competes have been outlawed, particularly in California, strenuously. So how do you enforce this? Because there are so many people who have moved from Apple to OpenAI. And there is at least a lot of allegations that they took information and they continued to take information. So has trade secrets like litigation become a replacement for non-competes for the really extreme situations? Yeah. Non-competes have become a lot more restrictive.

I mean, like, it's actually the non-compete, the litigation over non-competes in the past few years is, like, very fascinating and it's been very active. Like, you know, things have been overturned. I don't actually even know what the current status of it is, like these days, but they're either legal or not enforceable. So I don't know what the current status is. Like an appeals court may have overturned something like. California is unenforceable. Yeah. I think New York may be the same.

But anyway, that's not. Yeah, not the point. Yes. But it's interesting because like you're right. So with the open source conversation, does that mean that trade secrets become less important? Does it mean that like open source is a quote unquote better path? Because I don't know if anyone's ever been a trade secrets litigator or been involved in a trade secrets case. I unfortunately have been involved in the latter, not me personally, but just dealing with it for, you know, entities. And it's a hard charge to prove. And you,

You see, like, there's a lot of cynical people here because, you know, Hard Fork was talking about how, although the allegations are really bad, like they said something like Apple treats everything that happened in Cooper Tino as a trade secret. So like maybe they've been building this. But at the same time, like, we're just talking about one of the allegations here, which is like a guy who left Apple gets into the Apple's network storage, messages a friend who still works at Apple and says, this is a quote, LOL, I found out I can access the network storage so funny. And then she writes back, I'm ready. So look, could we interpret that in many different

ways? Not really. But that is the paragraph that gets you past summary judgment, you know, if you can show it. So it's really like interesting to look at this in the holistic view of these behemoths, Apple, Google, now anthropic, and Open AI, like they are competing to see who wins here. And litigation has become a tool in that. But that's what's so fascinating, right? Is that like if open source does start to become the default model, it really does change the nature of all of these fights and potentially like the law. And obviously like the need to litigate a lot of these things. So I think that's really interesting.

And again, I think like there are a lot of lessons to be drawn from crypto. I think we have actually stumbled onto a model that will prove more resilient. Right. So think about this, right? If your competitive advantage is open protocols, network effects, and interoperability, you just become a lot less dependent on protecting individual pieces of intellectual property or IP. Right. And you become like the need to enforce non-competes and like, you know, trade secrets and all of that just becomes a lot less important.

So I think like if open source does become the default with a lot of these technologies, it does transform the nature of all of those things. Okay. Anyway, let's move on. We are going to go to break first to hear a word from our sponsors. And then when we come back, our guests will be joining to talk clarity, defy lending, and vaults. So stay tuned. We'll be right back.

If you hold crypto on your phone, your biggest vulnerability isn't your wallet. It's your carrier. AT&T, Verizon, and T-Mobile have been breached again and again, and SIM swaps are still one of the easiest ways for attackers to drain accounts. That's where CAPE comes in, America's Privacy First Mobile Carrier. Same premium service, but CAPE rotates the identifier on your SIM every 24 hours, deletes your call and text metadata after a day, and protects against SIM swaps with a 24-word recovery phrase that only you control. You also get two middle-to-end encrypted secondary numbers for banking and sign-ups, so you stop handing your real number to every app that asks. Go to cape.co slash unchained

and use code unchained for 33% off your first six months. Joining us now is Salman Banai, the GC of Plum, also former head of policy at Uniswap and ChainLink, I believe, and also someone who was spend time. Oh, she analysis, sorry. And someone who has also spent time as a lawyer at both the SEC and the CFTC. And lately, I think you've been involved in effort to try to get clarity over the finish line. So we thought it would be great to have you on just to give us an update on all of that. And also Commissioner Persis' statement on defy lending and vaults, something that you and I talk about all the time just because, you know, that's what made and bloom is closely involved in.

And I think the statement could be best described as like a warning shot, but we'll get into that in a second. Do you want to start with a clarity update? I know you work a lot in particular with the Democrats in Congress and their staff. So I'd love to hear in particular like how they are thinking about it right now. Yeah, thanks, V. Thanks for having me on. Yeah, so I've been involved in some of the, this, the, the, the, the, discussions around clarity for several months. It's picked up last couple weeks. I've been on the hill several days over the last couple weeks, meeting with members and different offices on

both sides. I would say, you know, the sentiment is generally positive, but there's a lot of concerns that political factors may be affecting this legislation. that go well beyond, I think, the gaps on policy, which do seem more surmountable. But, yeah, I caught up with some lobbyist friends. Shout out to Keith Castaldo this morning to get the absolute latest. So you may have seen this morning in Morning Money, Ruben Gallego promised that he's working on ethics language with Tom Tillis. Tom Tillis being a Republican senator that's fallen out of disfavor with the president that's publicly, you know, called out some of the, you know, shady dealings of him and his family around crypto and other areas. So he's seen as a credible go-between Republicans and Democrats.

Gallego has promised, Gallego is a Democrat, by the way, from New Mexico, or from Arizona, I'm sorry. And he has been very pro-crypto since he got elected a couple years ago. And he really sees himself as one of the leads. He's the ranking on the Digital Assets Subcommittee at Senate Banking. And he's promising ethics language in the next couple of days. My understanding is that one of the red lines for Democrats is having some kind of state AG authority as a check on a Department of Justice that may or may not actually enforce the ethics provisions. So that's a red line, and I think we're going to see some language there, where the state AGs could have

some authority in enforcing the ethics provisions. I think that's what Democrats would like to see. There's precedent for this. The CFPB's statute extends state enforcement authority to state AGs as well as the federal government. So there is precedent for this kind of thing. I definitely want to get into the ethics provision deeper, but I have a question about something you talked about, which I know is obvious to you and maybe some of our listeners, but you talked about the difference between policy and politics, right, and how politics might get in the way. I think it might be useful to like dive into what you mean by that a little bit more because it's easy to sort of glaze over those comments and to people that don't know those words or don't work on the hill like

you do. It's sort of difficult to know where the winds are taking each part. Yeah. So it's really a matter of calculus, right, and goals. So when it comes to a political issue, it's the kind of issue that a member has to take back to their district or to their state and have to be able to defend themselves, not just from, you know, losing potential support in the general electorate, but probably even more importantly these days, it's the primary electorate. And so taking, you know, a position, for example, for a Democratic senator that could be seen as giving in to Trump could put them in danger with primarily a primary electorate. And so that's something that a lot of these members are very sensitive to. Given the dynamics, you know, there's been a number of,

you know, upsets in different primary races this year. That's a trend that's likely to continue. So that's what I call a political calculation, is like what is the cost benefit in terms of electoral politics. And then there are policy differences. So policy, a policy difference would be on a matter of public policy. So, for example, a lot of senators, are very concerned about illicit finance and the potential of if we bring digital assets into the mainstream, does this facilitate illicit activity, sanctions evasion, money laundering, human trafficking, etc. And so that is, that's something that reflects the senators' policy preferences. Democrats have some tendencies, you know, some priorities, Republicans have some

priorities. Oftentimes they might agree on the outcome, but they may not agree on the mechanics and the specifics of how to get there. And so that's that's that's that's a policy question. So one of the policy questions that's front and center is the extent to which defy and defy related intermediaries come under the bank secrecy act. So you know the the US is anti-money laundering statute. And so that's one of the policy debates that's happening. But this political issue around ethics has really overtaken most of the negotiations. A question on the ethics part, NBIA, I didn't mean to cut you off. But if I was an alien coming down to Earth looking at this, I would sort of think like, yeah, why shouldn't everybody agree that you shouldn't be enriched based on this, right?

and there's obviously different dynamics to it. But from the outside, it seems like this should maybe be the easiest part of the bill because it's less technical and trying to understand how to protect Defi and stop illicit finance and much more political, I guess. So I think it would be helpful to understand what the two sides are going for. I think some of us have a sense of what it is. but if you could lay it out like being on the ground rather than just getting the rumor mill. Yeah, so I think the, you know, the Trump and the Trump family have, you know,

engaged in different types of business activities, you know, since the president took office, I think his net value has gone up by, what is it, $2 billion. And I think a large amount of that came out of the meme coin issuance that came the weekend before he was inaugurated. That's really something new. That's not something we've ever seen before. And the use of a meme coin as a way of, you know, making money by a political official or as a mechanism to curry favor with the president is really unprecedented. And so Democrats' view is that should not be allowed.

We should not have these situations. At best, it creates an appearance of a conflict. and at worst, it's a mechanism to curry favor with the president and get favorable policy based on your holdings of beam coins that benefit the president's direct bottom line. And so that's really been the focus of this ethics language. Then there's just general business dealings, you know, including World Liberty Financial, which involves members of the Trump family and senior members of the administration's family. And there the concern is probably more general.

as in it's a business. It could be used as, you know, doing business with that business could, you know, curry favor with senior administration officials. But that's been done, you know, previously and the ethics restrictions that are currently on the books do not extend well into the family of a president or a senior administration official. And so that's a broader question. and that really isn't touched by the current ethics negotiations, but I suspect it will become a focus for oversight, particularly if Democrats take over one or both chambers next year. Yeah. And then what about the BRCA?

Is that issue pretty settled, or are you still seeing back and forth on that? I still see back and forth on that. What I'm hearing from a lot of Democratic senators and their staff is they want to see where law enforcement lands on this. If it ends up being a net positive for the broader law enforcement community, you know, the way this bill, for example, extends the regulatory perimeter to cover a lot of defy, including illicit finance, you know, while it's screening at the interface level or the defy messaging system pro level and the bill, you know, that is seen as a net positive. There is some concern that the BRCA carving out defy developers from having to register as money transmitters could, you know, hamstring law enforcement efforts. There's been a mixed public record among, you know, people in this administration, people in the last administration, as well as different law enforcement groups. It's notable that a couple key law enforcement groups have come out in favor of this bill as a net positive.

That includes the Fraternal Order of Police, which gave a statement, I think it was a couple days ago, saying that they support the bill, as well as the National Organization of Black Law Enforcement, which came out, I think, was last week in support of the bill as well. So there seems to be an emerging consensus among these law enforcement groups that this bill is a net positive. And so that could give cover to these Democratic senators that, hey, this bill, at least on the BRCA, DFI regulation front, isn't that positive? And something that they can work with and potentially support. Okay. Awesome. And do you expect to see a lot of action this week, like a lot of movement? I guess the, what is it, the August 4th or August 7th?

Yeah. So I think the goal is to get a vote on cloture. this is the biggest hurdle because you've got to get 60 votes. There are two Republican senators that seem to not support this bill, and one that is incapacitated. So that would require just doing the math. I think that's 10 Democratic senators to support. And we're really on the knife side as to whether we can get 10.

And I think it's really going to depend on this ethics language that gets negotiated. And this ethics language has three. key stakeholders. There are the Democratic senators that are leading the negotiations, so that's Ruben Gallego from Arizona, Kristen Gillibrand from New York, and then Adam Schiff from California. It's possible we could get something that Gillibrand and Gallego could support, but maybe we don't get Schiff. If we get Schiff, then I think that unlocks a lot of senators. Schiff has been, very principled and has a strong reputation on issues of. relating to government ethics.

And so him lending his support would, would, would actually bring a lot of credibility. And you could see above 60 votes potentially. Yeah. I would say, you know, there's probably about a third, maybe half of the Democratic caucus that is probably not going to support this bill almost no matter what. So the question is whether we can get above that that 10 Democratic vote threshold or not. Yeah. Okay. Awesome. Thanks so much for the update on that. I wanted to move on to Commissioner of Purse's statement on defy lending in vaults from last week. So Commissioner

Perce released a statement last week about vault and non-chain lending that I think is actually one of the more important crypto statements we've seen from her in a while. So I've seen a lot of takes on what this statement was trying to do. But I personally read it as like a warning shot, but also collaborative, right? She's basically saying, look, come in and talk to us because these things do raise regulatory questions and we want to better understand whether and how they may come within the SEC's regulatory perimeter. Salmon, how did you read it? Yeah, I agree with you. And I think one of the the key call-outs is for those of us in the industry to just come in and, you know, help educate the Crypto Task Force and the SEC staff in terms of the vault ecosystem. And I really break it down

And I really break that down into the, you know, what I'll call the vault supply chain. So both Vita and Plume are developers of vault protocols or, you know, on our side, it's the NEST protocol. So there's the development. Then there is the curation, which is developing essentially a vault strategy. You know, the vault will invest in these assets. It'll be automated. It'll be discretionary. And then there's what I call something in between, which is the vault administration.

So this is kind of the back office. So once the curator has designed the portfolio strategy for the vault, the vault administrator then carries out tasks like rebalancing the vault based on incoming stable coin and the overall strategy of the vault. So this is the vault administration function I see as analogous to fund administration, which is not generally regulated unless you touch on, you know, investment. advisor activity or transfer agency activity in the traditional finance context. VALCuration, though, you know, it does look like it could trigger aspects of securities laws, potentially commodities laws, depending on the portfolio strategy.

And then one of the things that Esther first also focused on was the application of the Securities Act of 1933, which governs registration, of securities. And so she, she, you know, pointed to the fact that certain vault tokens and the overall context of the function of the vault may trigger 33 act issues. Yeah. And she cited Reeves and, you know, our friend, every crypto lawyer's friend, Howie, as well. So actually, she doesn't say how she's, I think in the footnote, she cites Foreman. Does she? Okay. Which is like an option.

But yeah. But yeah, she does talk about investment contract analysis. But yes. So yeah, as you parse it, you have developing vault protocols. You have vault administration, vault curation, and then the nature of the token itself, the vault token, the receipt token. Yeah. I also think another thing that like people seem to be missing is that,

The statement wasn't just about Walts, right? It's also about the on-chain lending protocols. Right. Like she says in the statement, quote-unquote, on-chain loans, depending on the party's motivations, the plan of distribution, and other relevant factors, can bear the hallmarks of notes that are securities. So she's just basically laying out the READS factors.

So that has nothing to do with Wells. She's actually saying the on-chain lending protocols, like even direct interaction with them, can raise securities questions. That's right. Another thing, and this is something you and Jesse were talking about before I joined, is the extent to which there is a human factor involved or not and reliance on open source code as opposed to human discretion. And so that's also a factor that she points out. And she alludes to, you know, there may be certain structures that lack a requisite human element that might bring, you know, some of these vault-based products. out of the scope of securities regulation.

There's also another aspect that you and I have talked about, which is the commodities law aspect of this. And there's registration obligations for commodity pool operators that pool assets from investors and then invest in commodity interests, which includes not just commodity derivatives, but also options on commodities. I can't trust right now. I'm like the SEC stuff. The security stuff is like enough. I think you're you made a really interesting

point and one that does connect to the open source conversation of like human in the loop like how many times do we talk about that with AI governments. I would ask you guys like on the vault side, it seems like aside from AI, defy has been having this conversation forever like where is the human, how much control? And now it seems like each of these conversations are really being narrowed down. to different products, right? So now there's a whole conversation about vaults and what does that look like? And obviously, Agentic Finance is something I'm working on as well. And it's the same sort of rhyming of patterns across it.

Like, do you think that even though we haven't been able to solve sort of the defy holistic question of human in the loop, we're going to have to nail down on specific products? Yeah. So I think it's, there's an interesting common thread between what's happening here clarity, Hester Persis' speech, as well as a statement that FATF published yesterday on the application of AML on DFI. And all three of these, the common threat is control. When is there requisite control to trigger regulatory obligations?

And the way clarity and FATF, you know, if you squint your eyes across the two, is it's really control over the protocol. Is there a discrete group of people that have the ability to change the core functions of the protocol? And if there is, then some kind of regulation should attach. Now, what's interesting about the FATF report is it doesn't necessarily mandate the specific type of AML compliance that we've seen, which a lot of people at Defire complaining about this report. I'm focusing on the glass half full part, which is, that it does contemplate ways of complying

that don't necessarily require traditional, customer due diligence KYC requirements, and instead opens the door to some kind of on-chain native way of compliance and the form of blockchain transaction analytics, freezing and seizing capability, things of that sort. One really interesting shift. I've noticed in the last few years about how we talk about,

defy is like we've kind of almost moved away from this concept of decentralization as like the mental model and now it's this idea of control as like the organizing principle right and maybe decentralization was always a proxy of a kind for control but like I do think thinking about it in terms of control and discretion is a more useful way to to think about whether and how the regulations apply. Isn't it different? Like, is decent, something can be decentralized with many people sort of touching it, right? And having sort of human in the loop as we're talking about, right?

And so I sort of don't know if that's the right direction for us to go. I mean, like, look at us struggling with control over the past. I don't know how many years, right? And so I don't know what the right answer is. But maybe the better way to think about it is like centralized control. Like is there a centralized party asserting control? Right? Because like, you know, even with these open protocols like the on-chain defy lighting protocols,

there are actors that make decisions about how the protocol operates, right? So they will set, you know, they decide things like what collateral is acceptable and other things. like risk and oracles. And, you know, they make decisions that do affect how the protocol operates, right? But it's a lot of times it's done in a decentralized way. Like it's a doubt, I'm sort of like just like disperse like governance body. Right. So maybe the right way to think about it is not like decentralization is a proxy for control,

but it's a proxy for like the lack of centralized control. That's right. And that's that's the view articulated in clarity. Another interesting thread between clarity and the FATF report is that they both have carve outs for some limited control as it relates to things like cybersecurity. So, you know, there is this option of, for example, a security council, you know, voting very quickly to address some kind of cybersecurity vulnerability. And that in and of itself is, in clarity, it's especially clear, doesn't trigger, you know, the requisite centralization. That's like a carve out to facilitate that kind of cybersecurity mitigation. Yeah.

Yeah, I think that's something we want to incentivize, right? Without like then subjecting the protocol to like the full speed of, you know, regulatory. And then the next question after, well, if there is requisite control, then what kind of regulatory obligation should attach? You know, and my argument is that the control should extend to the kind of activities there should at least be an analogy between that level of control and regulated financial activity. And it's not necessarily, like if there is control, then the whole panoply of regulation attaches, it's instead tailored to the amount of control. And then what is

feasible within that, you know, scope of control from a regulatory compliance perspective? So, for example, you know, if they, you know, I'll use the clarity term, terminology, if there's a team that controls quote unquote a non-decentralized trading protocol, what is the extent of the regulation? Now, in the Clarity Bill, there is clearly authority for Treasury to regulate these entities as under a limited version of the BSA. It is a on-chain native way of compliance, so it looks a lot like what the FATF report suggests in its rather flexible approach.

So it doesn't necessarily trigger CDD K-YC obligations, but some kind of on-chain blockchain analytics would be appropriate under clarity. And that approach is mirrored in the F-report. Now one thing I was prepared to talk about, if I can just jump jump right into it, because I know we only have five minutes, is what happens if clarity doesn't pass, right? So a lot of, you know, people here in D.C. seem to think that the next Congress will not move forward on any kind of digital asset legislation, given that Democrats are likely to control one or both chambers. I actually think that there will be efforts to come up with

a new version of clarity. It probably won't be called clarity because of political baggage at that point. But the interaction of the Biden administration's approach with the Trump administration's approach and this underlying need to regulate the space and the recognition of that need that at this point is bipartisan, I think is going to generate enough momentum that there will be. Digital access legislation under the next Congress as well. There's different ideas around what it might look like. I think, you know, we, it will probably focus more heavily on ethics, you know, that'll definitely be a core part of it. And the, when you see publications like this FATF report, that also provides some institutional, you know, international institutional

backing for what an approach to regulating DFI would be. Now, I would expect Democrats to push deeper into the supply cane. But I do think that there is, and you saw this with the leaked version of the Democrats' negotiating points on clarity from early this year, where they had an exclusion for defy developers. It was just much more narrow than what ended up being in clarity. So within the framework of like this institutional architecture that's developing, you know, through reports like this VATIF report, which,

by the way, included representatives from Treasury in drafting it. I think we could kind of foresee where Democrats might land on future digital asset regulation. I think another area where Democrats will focus on is trying to more carefully avoid instances where it's easier, for example, for a corporate issuer to launch a meme coin and get around a lot of, you know, the securities regulations

as opposed to issuing, you know, common stock. So a lot to, a lot to work through there. A lot to work. We should probably cover the fad of stuff on another week. That's an important report that I think people are both reading wrong and is worth talking about too because whether this passes or not, but especially if it doesn't pass, that is going to be a model that people look to. For sure.

Thank you so much for coming. That's interesting that you think that like even if it doesn't happen now, that it's not like it's never going to happen. Although I think we can all agree that we do not want to go through this again. So still fingers crossed for like the next few weeks. Anyway, Salman, thanks so much for joining us. everyone KKK will be back next Tuesday. Can't wait and we will see you then.

Thanks, Fee. Thanks, Jesse. Take care. Thank you.