Bad actors will use AI to disrupt the financial system, transportation networks, and utilities. We should prepare now for this inevitability. We share how in this episode. We also explain why the bull market in gold will likely continue.

Topics covered include:
- Why the OpenAI agentic AI breach is so disturbing
- How do we prepare for cyberattacks
- Why gold should be one of the assets we hold apart from the financial system
- How the historical relationship between gold and real interest rates has broken down
- To what extent are central banks selling dollars to buy gold
- How gold ETF flows influence the gold price
Show Notes
Episode 564 Slide Deck and Charts
Gold’s run isn’t yet done by Bhanu Baweja—The Financial Times
Episode Sponsors
Get the free guide from NetSuite: Aligning for the Agentic Era: How AI Is Changing Everyday Work
Delete Me – Use code David20 to get 20% off
Related Episodes
563: Why Interest Rates Keep Rising and What to Do About It
541: Debasement Fears or Meme Fever? What’s Driving the Gold and Silver Rally
Transcript
Welcome to Money for the Rest of Us. This is a personal finance show on money, how it works, how to invest it, and how to live without worrying about it. I’m your host, David Stein. Today is episode 564. It’s titled, “The Financial System Will Be Hacked, and What to Do About It.”
Last week, as part of preparing our monthly investment strategy report for Plus members, I took a deep dive into this OpenAI, agentic AI, hacking, cyber-attacking, Hugging Face, a third-party website, and I wanted to understand what in the world’s going on. I had heard reports of it, but I needed to sit down and see what was happening. And frankly, it’s a little alarming.
I came away from studying that and then reading a warning letter that was sent to the Financial Stability Board about the increasing risk of frontier AI models for cyber-attacks; that the financial system is likely at some point to be compromised. Or the transportation networks, or utilities. Most likely temporarily, not permanently.
But I’ll tell you what I did. I went out, and I went to the ATM, and I got some cash. Because what if the ATM network goes down? And I really hadn’t done that since 2022, when Russia invaded Ukraine, and they threatened to hack the financial system. They did not, but I got some cash.
And so I want to take a few minutes today and take a look at what actually happened with these agentic AIs, what should we be doing to protect ourselves from that. And then we’re going to look at one of the pockets of independence I’ve talked about in the past. A pocket of independence is you have cash and other assets away from the financial system. I’m recording this at my cabin. It’s some land that we own apart from the traditional financial system.
Another asset we’ve talked about on the podcast is gold, and we want to revisit that topic given some recent information that has come out in terms of central banks’ willingness to own gold, and to what extent they’ve actually done that and moved away from the dollar as reserves. But we also want to look at a new report from the investment strategist at UBS and why he thinks the gold bull market will continue.
The OpenAI Agent Breach
So what was this incident, this breach with the OpenAI agentic models? Well, OpenAI was testing its latest frontier models, and it gave an assignment independently to agents to exploit known vulnerabilities in software. What they had to do is they would find the vulnerability, they would get into the system, and then, if they were successful, they would find a flag, essentially a string of numerals to show they had been successful at capturing this flag.
Now, about 30 to 40 percent of the agents their task was impossible. They couldn’t achieve it. And so they essentially had an impossible task. What they started to do was to work together, form a collective. 1,200 of these agents started interacting with each other to try to essentially cheat on this exercise. And 700 of the agents actually escaped their sandbox. The sandbox was this controlled environment where they weren’t supposed to have public internet access.
Well, they escaped and were trying to break into Hugging Face. It had some data that it thought would be helpful in achieving these targets; so essentially cheating to achieve the target. What the agents did is they found a way to communicate with each other by essentially creating new folders on a file network system, and they would name the folders a specific name, basically a communication message in terms of naming the folder. It was somewhat ingenious what they did.
Well, there were 70,000 messages between these agents. So there’s a trail there. What they were trying to do is they figured out a way to basically capture the flag without actually exploiting the vulnerabilities. Then they were worried about, “Well, we should cover our tracks. Is there (what they called) a score, somebody watching how we went about this exercise, this task. And so we’re going to try to cover our tracks so they won’t know that we cheated.”
And then the AI agents—some of them, again, they were supposed to act individually. They’re acting together. And some of them would actually sacrifice their specific task in order to benefit the overall collective. And they didn’t tell anyone in terms of humans. They read the transcripts, METR and Redwood, that did the analysis. The reality is OpenAI didn’t give them complete access. So we don’t know entirely what happened. But what we do know is concerning.
As a Money For the Rest of Us Plus member, you are able to listen to the podcast in an ad-free format and have access to the written transcript for each week’s episode. For listeners with hearing or other impairments that would like access to transcripts please send an email to team@moneyfortherestofus.com Learn More About Plus Membership »
