JPMorgan's Dimon Sees Benefits From AI, Warns of Risks

Author: Bloomberg Television · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: JPMorgan’s Dimon Sees Benefits From AI, Warns of Risks · Channel: Bloomberg Television · 04:45 · Watch on YouTube

Topics: AI agents, Cybersecurity and software security, AI governance and regulation

Summary

Jamie Dimon says AI can bring major benefits, including better services, scientific advances, and fewer road deaths, while noting that JPMorgan has used AI since 2012 in areas such as risk, fraud, and underwriting. He warns that AI tools can expose serious vulnerabilities in software and that many industries, vendors, and public institutions are poorly protected from cyberattacks. Dimon argues that industries must help secure their own systems and vendors, with government coordination, and describes additional risks from persistent AI agents and agents authorized to make bank payments. He says action should come first, with debates over regulation and legality addressed later.

Key points

Dimon says he sees both substantial benefits and serious risks from AI.
JPMorgan has used AI since 2012 for tasks including risk, fraud, marketing, underwriting, design, and customer service.
He expects AI to contribute to advances in areas such as disease treatment and materials, and to help reduce road deaths.
AI tools can uncover significant vulnerabilities in software that organizations believed was secure.
Dimon says some industries, government institutions, and hospitals are poorly protected, and vendors can also be a weak point.
He urges companies to secure their own code and vendors and to help smaller businesses in their industries.
He describes industry-based efforts to protect critical infrastructure, alongside a coordinating role for government.
Persistent AI agents and agents with access to bank codes could create new risks, including unauthorized payments.
Dimon says organizations should address the risks promptly, leaving debates about regulatory models and legality for later.

Chapters

Start	Chapter	Summary
00:00	AI’s potential benefits and risks	Dimon says AI can be harnessed for good and describes its existing uses at JPMorgan, while acknowledging that it also has a harmful side.
01:35	Cybersecurity vulnerabilities	He warns that AI can reveal serious software vulnerabilities and that many industries and their vendors lack adequate protection.
02:36	Industry responsibility for critical infrastructure	Dimon calls for organizations to address AI-related risks, secure their own systems and vendors, and coordinate with government.
03:41	Persistent agents and banking risks	He discusses agents that continuously seek to attack systems and the dangers of granting AI agents access to bank accounts and payment capabilities.

Quotes

I’m in both. — 00:31

We’ve been doing it since 2012 — 00:31

I don’t think it’s existential threat — 02:36

The industries have to do it. — 03:08

get at it or about the legality of it later. — 04:45

Concepts and entities

AI, cybersecurity, open-source code, persistent agents, JPMorgan, Jamie Dimon

Full transcript (original language)

00:00 Which which end of the spectrum are you on? Because there. There are people who do and do absolutely believe that we are we are hearing that. Then there’s the other side that says we can harness it for for good. It can be controlled. It can be harnessed for good. Where are you on that, on that spectrum? So first the this is instructive of our bureaucracy, how it grows over time. Stupid things. That recession was supposed to be a summary of a prospectus, and that was the 80 pages on their own. I mean, and this happens all the time

00:31 and stuff. They just grow over time. I’m in both. So the big context eye is real. We’ve been doing it since 2012 that all of a sudden, uh, we use it. There’s thousands of case that we use it for. We have thousands of people do it, use it for risk, fraud, marketing, underwriting, design, customer service, you know, extra error rates as you’re going to see more and more in a good way. It’s going to, uh, save a lot of lives like you have, you know, in this today we have my view is of probably will literally the next 10 or 15 years save a

01:01 lot of diseases Cancers have new composite materials. Your airplanes, your engines. You know, 45,000 people a year killed in the streets and your 150,000 crippled. That number could be cut by 80%. So, you know, and I think the generation, you know, 30 years now, they’re probably working for 3 to 5 days a week. They’ll have an income of per person or $300,000. It’ll be a great life. I just that’s the good part. And I think there’s truth to that. And, uh, it’s also true that there’s a bad part, and I’m not going to the jobs part unless you want me to, but but the

01:35 bad part is the risk that some of these things now pose on creating. And you all have systems that use applications. We use 7000 applications, a lot of open source code. There are hundreds of thousands of open source code. And these things are so powerful now that you thought you did, had no vulnerabilities. And you run this through your own code and there are hundreds of significant vulnerabilities. And so for JP Morgan, you were we spent a lot of money and cyber was very well protected.

02:06 You’re probably up there with the best in the world, and a lot of the banks are, but there are industries completely unprotected. You know, you guys are probably pretty well protected, but then you have vendors who aren’t. Absolutely. And their industries are like water zero, which are government institutions zero. Sure. And hospitals zero. So they are hugely valuable now to state actors, terrorists, uh, criminals or just people who hate you, you know, who don’t like the company. They don’t like the industry, like the party, they don’t like the, uh, and we have to get our hands on that.

02:36 And where that risk is. And how bad is it? I don’t think it’s existential threat, but, uh, you you’re going to see problems, too. Let’s roll up our sleeves, which we’re doing. We have a lot of people doing it out to fix the problems, so that we actually are on top of what those risks they create. So we get the good and not the bad, but it is critical. And the government is geared up for this show. We, Jay Clayton, who some of you probably know was just announced the, uh, czar for the government, which he’s the national director of intelligence. We’ve been in contact with him. You know, they all know the risk. The government organizer.

03:08 I didn’t know how to do it either, by the way. So it isn’t like they’re sitting there saying, we’re on top of this. They’re saying we need help and the help is us. They can’t protect all these industries. The industries have to do it. And Scott is on this thing with me called the lines for Critical Infrastructure. The CEO of a united, uh, which is trying to do it by industry so that every industry should have its own. Protect yourselves. Clean up your own code, clean up your vendors codes, help the smaller companies in your industry like we have, you know, thousands of banks that we bank who are quite concerned about them. And I am too, you know, and, uh, and then you have different vendors and, you

03:41 know, they’re the way you look at there, 20 industries. You know, we’ve started with six, but we go to the other 14, and all these companies are putting real muscle into at this point, and the government is going to organize around it. Uh, but we have to do in due haste. And this is one we have a common interest of China, because I think China realized this isn’t about us versus China. In the in China, they also have to worry about people like the government, terrorists, state actors don’t like China. Uh, if this could be used against anybody. And then it’s made worse, by the way. But these agents try to get to me by the

04:14 agents mutual instinct and all that. So now this I created, you know, a little of what we call persistent agents that they’re always looking to attack you. And 24 over seven. And this agent talks that agent and that agent, first agent and, and, uh, and if you allow them in your bank, this is a big issue for banks. You know, if the muse is trying to get they can make payments for you. That means you have to give them your bank code. That means they’re in your bank, and you’re assuming they act properly on your behalf, and they may not. And of course, they may do things inside

04:45 a bank, which is treacherous for the bank, too. So these are big deals, and I don’t know how we can deal with them. All of them. Robert. So if you get to work, the government will obviously play a role. It was just a big debate about, you know, it should be like the FDA, a government regulated thing or like a Finra, which the securities self-regulating thing or just industry getting and getting that. And I don’t really care. My view is get at it or about the legality of it later.