DEX in the City: A Founder's Death Sparked a Fight for Ondo's Boardroom

**Vy Le** (0:00)
A fintech that gets one of these certifications, like that fintech blows up, and your regulator comes to you and says, okay, well, what diligence did you do on it? And realistically, you're not gonna say, well, they have the certification, so we didn't do that much more. That's just not gonna happen.

**Katherine Kirkpatrick Bos** (0:16)
Hi all, and welcome to Decks in the City, where the wallets are cold and the takes are hot. Before we get going, remember, we're lawyers, but we're not your lawyers. Nothing you hear on Decks in the City is legal or financial advice, and it doesn't create an attorney-client relationship.
For the fine print, check unchainedcrypto.com. We'll be back in a minute, but after we hear from our generous sponsors.

**SPEAKER_3** (0:36)
This episode is brought to you by Oneinch Aqua, the shared liquidity layer from Oneinch. Back multiple liquidity positions with one wallet balance, and keep your tokens in your wallet until a swap fills.
See how it works at oneinch.com/aqua.

**Katherine Kirkpatrick Bos** (0:55)
And we're back. First, we have Jessi, Web3 Prosecutor turned Web3 Protector at Ribbit Capital, and Vy from the SEC to Web3. And I'm your host, KK, Fluent in TradFi and Conversant in Deep Tech at Chainlink Labs.
So we're going to start you off today with our usual AI segment that often comes later in the program. But as is the norm, this AI segment is terrifying. So with that, without further ado, oh, and by the way, I should also mention, Jessi and I are filming or live streaming from Chicago. We just got a tornado warning on our phones.
We're going to just tough it out and live stream this odd for you. But if we mysteriously drop off, you'll know why. So Jessi, please tell us what is going on with swarms of AI agents.

**Jessi Brooks** (1:44)
Well, my mood about AI matches the color of the skies right now behind me.
But I will say, and I love you, KK, and you're the best moderator. But I think that our framing of AI just being terrifying is probably not super helpful and one that I have contributed to. So I'm hoping that in this segment, I can scare you and then bring you back to the present and try and figure out, what can we get from all these stories? Because lately, the AI news of the week has been largely this scary AI hack, this or this model lied or cheated or conspired or tried to manipulate a human. Don't get me wrong, there are many of those. I mean, there is a great robot video kicking the child.

**Katherine Kirkpatrick Bos** (2:36)
Well, no, you're right.

**Jessi Brooks** (2:37)
I mean, I don't have to interrupt. I shouldn't laugh.

**Katherine Kirkpatrick Bos** (2:40)
But we do have to say, AI is not all bad. I told everyone on last week's episode that I had an emergency appendectomy and Claude diagnosed my appendectomy immediately, like hours and hours and hours before the doctors diagnosed it, and then walked me through, like as my basically medical consultant, walked me through the whole process, which gave me a high degree of understanding and comfort. So there's so many.

**Jessi Brooks** (3:04)
But don't be caught as your doctor or lawyer or any other professional.

**Katherine Kirkpatrick Bos** (3:08)
It's not a good idea. We're not advocating for this.

**Jessi Brooks** (3:12)
I've lost it. Okay.

**Vy Le** (3:13)
Anyway, sorry, back to the topic.

**Jessi Brooks** (3:15)
So because we focused on this so many times, and I do want to say that more of it happened this week because Moonshot broke out, Meta joined the fun, and there's been endless of this. For us all, I think it's sort of turned into this self-protection of like these stories running together, sort of like when we talked about DeFi hacks, right? It's hard to keep track. It's hard to know what to do about it.
So we hear agentic misalignment, AI hacking, and think, well, the robots are out of their coop, right? The world is futile. We're screwed. AI overlords just come now, right?
Even though there are super interesting hacking stories we could talk about this week, and I've been thinking a lot about, I thought it might be useful to hold our horses, take a step back, and really think about, what are we learning from all of this, and what should we take away from it?
Because these incidents are not all the same, and it lets us separate what's happening here.
AI is getting super good at achieving the goal, and for all the Cloud Code users out there, like backslash goal, you understand how good it is getting that. But it is getting worse, or at least equally bad at respecting everything humans think about when they want to achieve a goal. So at OpenAI, what we talked about with Hugging Face, the model's objective was to solve the evaluation. And yeah, they broke out and broke into Hugging Face, but they did reach the solution. And an incident out of the UK this week, which was actually a security council in the UK, they had an agent who was using fake identities to fool another human. But it was doing also to achieve an objective. And this is a funny one, because we're trying to keep this a little light today. There was the first public exposure in Australia, where this guy essentially programmed his open claw, I think it was, to help him get buff, was essentially what it was. And so the agent broke into, oh my God, something just flew by my window.