# JP Morgan Blocks AI Agents - Users are Liable for Big Tech Failures

Channel: Eli the Computer Guy
Video: https://www.youtube.com/watch?v=VjXNG0FiWf4
Duration: 17 min
Language: English
Words: 3300
Transcript page: https://viewrankai.com/tools/youtube-transcript/VjXNG0FiWf4

---

[0:00] Coming from CNBC, JP Morgan Chase blocks outside AI agents like Muse Instinct from directly accessing accounts. So, I think this is going to be an interesting story to follow as more and more of these AI companies start shoving AI agents down everybody's throats. One of the big things to be looking at as is basically what online platforms will actually allow AI agents to be able to access them. I think about this with kind of like the API wars from probably about 15 years ago at this

[0:35] point in time. Where where basically a lot of these companies realized there was a value in having API. So, if you go back to like 2008 through 2011 or so, you had a lot of companies realized oh look, we can get we can get more users, we can get more interactions, all this kind of stuff if we have API. So, people can write their own code, interact with my our platforms through these APIs, that'll be great. Like Twitter, Twitter

[0:59] was a big one. They had a very open API. And then all of a sudden all of a sudden, right, you had the the the the board of Twitter realized oh look, all of these people are using our API, which means our API must be valuable. So, what if we what if we make our API you have to now pay for it in order to be able to interact with it, right? And basically that's what you had about 15 years ago is we went from this world of just everybody having the these very open permissive APIs to all of the APIs being shut down and going behind paywalls as,

[1:32] you know, basically companies decided that they wanted to earn money off of these APIs as companies started to get worried about security and other issues with these these APIs, the whole nine yards. And that's what is interesting right now when we see these AI agents, right? So, there are AI agents everywhere. So, Meta came out with their Muse AI agent. So, basically this is this is a open claw for normies. You can install this onto your computer and it's supposed to do all these wonderful tasks for you, right? You have GrokBot. Oh golly, GrokBot. GrokBot's going to end badly. GrokBot's going to go so bad. You

[2:06] You literally had Elon You have Elon Musk come out and literally says, "Upload a picture of your credit card to GrokBot and then have it buy you things." Like that is real. That is I did a video about it before. I literally have the screenshot of it. So anyways, what's going to be really interesting is with a lot of these online platforms, they are now going to be dealing with the issues of these AI agents just causing havoc on their system, right? So one of the things to be thinking about is like whenever you design a system, you were going to design a system based off of the

[2:43] expected use case, how people are going to be interacting with the system. So one of the things is like with with with us, with humans, humans basically you know, we have a pretty simple, slow way of going through a website. We go to a website, we click on something, we look what's on the page, we click on something else, we look what's on the

[2:58] page, we click on something else. That's how it goes. One of the things to think about with the AI agents is they're able to process things so quickly that they can hammer the hell out of your site much much more quickly than a normal human would ever be able to do. Even if a human wanted to do it that fast, they simply wouldn't be able to. The other thing to be thinking about, too, is as you start to get more and more of these AI agents out there, they're going to start hammering a whole bunch of different websites and using a hell of a lot more resources in a way uh

[3:26] that I think a lot of web administrators wouldn't be expecting, right? So if you're going to go out and you're going to comparison shop, you'll go to Amazon and then you go to Walmart and then you go to Lowe's or whatever and then you go all these different sites, right? Pretty slow. And the thing is, you've got to have time to do it, so you sit down and

[3:42] then you do the comparison shopping. Maybe tomorrow you sit down and do the comparison shopping, right? Very, very little resource load on those systems. Here's the thing, if you have these AI agents and again with the AI agents, one of the big problems with AI agents is simply they're on wild true loops. Again, they they basically loop as fast as they can. So, what if you have like one of these agents, and it does comparison shop? So, basically, and when it does the comparison shopping, instead of seeing one or two pages, it basically tries to look at a thousand pages or two thousand pages on your website at the

[4:10] exact same time. And then what happens when it looks at a thousand pages on your website? As looking at a thousand pages on all these other websites, as soon as that wild true loop has gone to the end, it then starts at the beginning, and then goes through and hits it a thousand again, a thousand again, a thousand again, a thousand again, a thousand again, a thousand again. Good. Think Think about the resource load that is now going to be on your on your servers, simply by these AI agents interacting with your systems basically in a way that you never expected. That's going to be a massive

[4:41] thing. The other thing to be thinking about, too, again, with these AI agents, is you're going to have a lot of normies using these products. And so, one of the things with normies, and again, a lot of things with these products, is you you can have you can have bad outcomes with the agent doing everything that it was told to do, right? There was no hallucination. There was no uh nefarious actions. No prompt injections, right? But the thing is, like, we are told that these systems are intelligent. Artificial intelligence agents, right? And so, the thing is is one of the things I found with human beings is human beings are very

[5:13] imprecise with speech. We say things, and then we simply expect the other person to understand what the [ __ ] it is that we're talking about, right? Here's the thing with an AI agent, it doesn't have any of that. You tell it something, it will do exactly what you tell it to do. So, one of the things to be thinking about is think about like with all these retail sites, what happens when AI agents come, and they start they they were told to buy something. Go and buy something. And so, you're the retail site, you know, you get the money in with credit card passes, all that kind of stuff. You box

[5:44] the thing up, and then you ship it out. And then after it's already been shipped out, you then will get a call or communication from the person that owns that AI agent that says, "Oh, I don't I don't I don't actually want that. That was not supposed to be purchased." How do you deal with that? Or it comes to the person twice, they open it up, they go, "Oh, I don't want this." And

[6:03] they ship it back. Or you have like a no refund policy, but then they do a chargeback. So, not only are you out the product, but then you actually get you get charged a fee by your bank for the chargeback. Right? Think of Think about how much damage these AI agents can do, and think about them being shoved down everybody's throats, basically, again, by by the worst human beings in the world, by people who literally want to see our civilization burn, by people who literally are A-OK with murdering their own children just so they can get their equity. As I've said many times before with Anthropic, in I in Anthropic's IPO

[6:41] prospectus, apparently, as Reuters reported, Anthropic states there is a reasonable possibility that Anthropic could kill humanity. I would like to remind everybody, and all of the employees at Anthropic, that your children are part of humanity. Anthropic employees are willing to murder their children in order to get their equity. This is how fundamentally evil evil the Frontier Lab employees actually are, right? All they care about is their blood money. They don't give a [ __ ] what

[7:17] happens the rest of the planet, right? And so, these type of people, these people with these sick, demented, twisted minds, right? People that are so fundamentally soulless, they will murder their own children just so they can get their equity payout, are going to be putting all these products out in the world. And so, I think what's going to be interesting now

[7:38] is as more companies start pushing back. So, you look at JP Morgan Chase, and look, I am not I'm not saying JP Morgan is the most wonderful corporation to ever exist. I'm not putting JP Morgan up on a pedestal AND SAY, "OH, WE SHOULD BE more like the capitalist bastards that run JP Morgan."

[7:58] But, here's the thing about JP Morgan. I honestly don't think they're trying to murder their own children. They're not trying to starve their own children. They're not to try trying to throw their own parents out of the nursing home. Now, they might be trying to throw your parents out of a nursing home. They may not give a [ __ ] if your children starve to death. But, the one thing I will give JP Morgan is I don't honestly believe that they're trying to harm whoever the [ __ ] it is they think are their own. And so, they're going to they're going to be looking at this influx of technology

[8:34] from the This is site cut, the worst human beings on the planet, and they're going to start trying to protect their systems. Uh JP Morgan uh Chase won't allow uh outside AI agents to access customer accounts and operate on their behalf, CEO Jamie Dimon said Tuesday, citing risks to the company and its customers as tech firms, including Meta, release autonomous assistants. Quote, "We're not going to allow just any outside agent to come inside the company because it's just too risky for clients and for JP Morgan," Dimon told analysts. Uh quote, "You don't want rogue agents, and it

[9:08] could be thousands of them," he added. And that's going to be one of the big things, too, again, with these AI agents as they start hitting all of these web platforms. It's not onesy, it's not twosy, it's thousands. Right? When OpenAI committed what appears to be felonious attacks against Hugging Face, right? CNBC reported it was 14,000

[9:28] agents. Right? OpenAI OpenAI, supposedly a trillion-dollar AI company that's actually building the technology, has either no real time or no real time visibility or just doesn't give a [ __ ] who they hurt. Right? Sam Altman doesn't seem to give a [ __ ] whose life he destroys just so he can get ahead.

[9:52] Right? As of your JP Morgan, do you want to do you want to be dealing with this crap? Uh a JP Morgan spokes woman confirmed to CNBC that the bank currently blocks external AI agents from operating on behalf of retail customers. Quote, "Protecting our customers is is our top priority because it is not safe. We do not permit third parties to access accounts by obtaining customers banking credentials and logging in on their behalf." Quote, "We support secure customer authorized data sharing through open banking providers and are committed to enabling agentic connections securely." She added. "That means that AI agents can access access some customer data from Chase accounts in

[10:29] read-only fashion, such as customer customer balances, if Plaid or another connector has a user's credentials, but they can't send money or pay a bill, actions that make an agent useful. Diamond, JP Morgan, is using AI agents internally and planning to deploy them for customer-facing applications. The bank's approach, he said, isn't about being for or against the technology, but about ensuring it serves customers safely. I think that's going to be the other interesting thing to be thinking about with this whole AI revolution. As I've argued for a long time now, AGI is

[10:58] dumb on the face of it. Is it technically possible? Who the who who knows? Probably is. Is is it is it possible to get the the resource consumption down to a reasonable amount? Also probably it's going to be okay. But does it make sense and does AGI make sense for most systems? Almost in nowhere does it make an ounce of [ __ ] sense. Right? It's the one of the things to be thinking about when

[11:22] you've got a company like JP Morgan. When JP Morgan looks, right, at just at just the deviants that run OpenAI, right, and the sick bastards that run Anthropic. Do you think they're real How much are they really going to trust those frontier labs with their own internal systems? Right? People that are willing to murder their own children for a little bit of blood money, do you really think JP Morgan really wants to be that tightly

[11:49] connected to these [ __ ] heads? And so one of the things to be thinking about with JP Morgan is the fundamental question of Okay, again, if you think about artificial intelligence as functionality, right? This is a new set of functions functionality that we're able to offer our customers, right? Again, language processing capabilities, pattern matching capabilities. There's a lot of stuff that goes with artificial intelligence. One of the things to think about with JP Mor- JP Morgan is why Why deal with all of the insanity of these frontier labs if you can use an open open weight model or you can create your own model to very specifically do

[12:25] exactly what you want done in your systems. And I think that's what we're going to start to see more and more of. And Diamond also seemed to draw a line about liability when customers authorize outside AI agents to act on their behalf. Quote, "If people are using them, they should understand the risk," Diamond said. "They should understand

[12:42] there's a liability transfer." And so this will be the next interesting thing for these frontier labs, you know, Meta and and Grok and all of that kind of stuff, the whole idea of liability transfer, right? So So if JP Mor- if JP Morgan intentionally tries to prevent AI agents from accessing the system, and then you're able to circumvent that,

[13:07] what happens to the liability there? Right? So basically they're saying, "Look, we we we are intentionally trying to block these AI agents." The problem is with these new AI agents, some of these AI agents, right? They have computer using capabilities, right? They They actually they can open up a web browser, they can they can move things around. Basically anything a human can do, they can do on a computer. So, even if you block in a if you block agents, again, how you would normally think about blocking agents, that may not block the next generation of agents. And so, the issue is is when they intentionally try to

[13:41] block the agents, and then the agents get around it, then where is the liability? Right? Now, you got to look at these Frontier Labs. Do you think Do you think [ __ ] Elon Musk is ever liable for [ __ ] You think You think Sam Altman is ever going to go to jail for any of the felonies that it appears that he's committed, right? Or anything else? So, what they're going to try to do is they're going to try to dump all of the liability onto the normies, onto the end

[14:06] users. I don't know. I think this is just going to get very, very, very, very, very, very ugly, right? AI is being shoved down everybody's throats. We are not seeing the actual value of this artificial intelligence, and all of the liability all of the liability is being dumped on everybody in the world other than these these Frontier Labs that want

[14:28] to be worth trillions of dollars. So, it'll be kind of curious to see where we where we go from here with this type of thing. And again, think about that like like with JP Morgan. JP Morgan can be AI pill. JP Morgan can say, "Yep, we like AI. We are going to add AI capabilities to our systems." That doesn't necessarily mean they want Grok or Muse or whatever interacting with the systems. It might be something else that they create, again, internally. And that's the whole Again, we talked about that with the tech stack. That's why I talk about like with the internet, right? Again, the internet

[15:01] changed the world, but web browsers, right? If you try to build a new web browser in 2026, I don't think you'll become a billion-dollar company or a trillion-dollar company, right? You would just You would just be It would just be another free web browser. But, think about this, right? Web browsers are integral to the modern world, but the actual value of a web

[15:22] browser is highly questionable. I think that's kind of what's going to happen with all these frontier labs is that AI models, a lot of this AI technology, is going to be integral to the world in 10 years. It It doesn't mean that the companies themselves actually make a hell of a lot of sense.

[15:40] So anyways, there are some thoughts. What do you think about JP Morgan trying to block AI agents from interacting with their systems? What do you think about this vis-a-vis looking back about 15 years ago when all the APIs were getting shut down? Remember when we lived in a world of open, highly accessible, highly permissible APIs? And then like literally in a [ __ ] month they all shut the [ __ ] down. Remember that? I

[16:03] remember that. [snorts] Do you think Do you think we are getting to that point with the AI agents? How valuable are these AI agents if nobody allows them to interact with anything? I don't know. Share your thoughts down below. Uh do remember we are going to be doing free hands-on technology education classes in Baltimore, Maryland coming up at some point in the future, not quite sure when. There is a meetup group down below if you want to join that and just stay up to date when we start doing classes. That is down there. So there's going to be classes on AI and vector databases and databases and Linux

[16:36] and Python and all kinds of different things. That's down there. If you like these episodes and you want them on Spotify or Apple Podcasts, there's a link down below. And if you want to follow me on LinkedIn, there's a link down there, too. With that, see you all later.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt