Discussion
I just got a letter from my bank saying "do not allow AI agents to have your payment information for any of your accounts."
Full stop. Just don't use any agents. It's dangerous.
It didn't contain additional language saying they won't protect you if they do, but I wonder if that could be next.
I wonder if other banks are doing the same.
Maybe they will ban them.
It seems reasonable that, if an agent makes a financial transaction, it ought to be identified as such, no?
I don’t see how “agents” can be made reliable or that there can be an extra flag or something to inform processors of transactions that are initiated by “agents” - not practically at any rate. There’s just the system for processing payments that we have. If one party has all the necessary authorization and the transaction isn’t fraudulent… why shouldn’t a bank honor the contract?
When a person gives an “agent” their payment information and authority they have made a contract to throw some dice and pay depending on how the dice land. I know that millions of people rationalize away this logical implication of what they’ve done but that doesn’t change anything.
You could have alternate "agent only" accounts with extra insurance fees and help with credentials to be given to agents. That would seem like the minimal prudent action if this catches on at all.
But the frostiness from banks makes me hope that it won't.
Anyone who gives Mark Zuckerberg their bank or credit card information is an idiot.
Then they number in the tens of millions, facebook marketplace is enormous.
So ready for AI and Social Media regulation.
@futurebird pretty sure my accounts have a caveat of “you aren’t covered for loss if you give your password to anyone else” so giving it to the bot of some company would qualify, imho. So they have already said they wont protect us? The letter was just a helpful hint ;)
@futurebird They already have language that says you won't be protected. If you knowingly give access to your bank account and all your money is taken then it's all on you.
Whether you're giving access to your ne'er-do-well son, the guy you just met at a bar or an "AI" "company" doesn't really matter. You granted permission. You're responsible. That's what your contract with the bank will say.
@futurebird or banks will introduce their own agents to do it “safely”?
@futurebird "It seems reasonable that, if an agent makes a financial transaction, it ought to be identified as such, no?"
In my opinion, we should have an international law that AI agents must never pretend to be a human and must always truthfully respond when asked if they are AI. I know regulating this new technology is difficult, but this one is a no-brainer for me. Bad stuff will happen if we don't enforce this!
@futurebird
Wells Fargo has built in AI on customer apps that is an impediment to looking up past deposits and payments. I’m going to local branch in person tomorrow to see if a human can show me how to use it, grr.
Kind of wild that a bank is saying "do not use this service from a major company"
But thinking about it they are going to be miserable from people trying to undo agent mistakes and mischief.
More darkly I wonder if they want to be the ones who run the agents and it's just a market grab. As if people want this mess.
@futurebird Agentic Crime is kicking-off and likely to get a whole lot worse. There's so much potential for financial fraud and theft at scale.
I think the bank's message is for a combination of reasons:
- Customer education so people have some degree of self-defence against scams.
- The bank's Customer Support department might already be getting tired of dealing with the fallout of people who think "I've done my due diligence" and gave control of their banking to stochastic parrots. Not all LLMs are run by people with malign intent; many are run by people who are ignorant and arrogant and in a rush. Regardless of motivations, no LLMs are intelligent, so serious errors are inevitable, therefore giving LLMs access to your money is certain to lead to financial disaster. It's not a question of "If" but "When".
- The bank's Legal & Compliance department wants to say they attempted to help and it's not their problem: plausible deniability.
I'm very happy for this notice from my bank as mentioning that "banks have warned about using AI agents" gives my little warning article for the faculty newsletter just the right extra "umph"
I also mentioned a "robust debate about the safety and utility of AI agents among security professional online" so that's ya'll's bit in this one.
I had to make it "neutral" ... we'll see if they publish it.
I do sincerely think you should warn off all of your friends against using any kind of agent. It's dangerous. I don't have confidence the people who get cleaned out will be made whole. None at all really.
And there are some kinds of damage, information leaking damage that cannot be fixed with all the money in the world.
Call your mom, call your aunt.
"Hi, Uncle Fred?"
.....
"Yeah, he's doing well. How's Nancy?"
....
"Good, good. Listen, Fred, I just wanted to let you know that if you give your bank account info to some AI agent, like that Muse thing, and you lose all your money?...... Well, if that happens you can't come live with us."
......
"Yes, I agree, family's family, and otherwise, sure. But lose it all to AI? That's just too dumb. It'll be a 'nope'."
......
"Okay, we clear?"
.......
"Good. Love to Nancy. Bye-bye."
For me the warning from banks ... well it is nice, but it's not what convinces me at all.
@futurebird As long as just humans are involved, there’s only a finite amount of dispute resolution a bank has to worry about. Once AI agents get involved, the disputes needing resolution increases by magnitudes of order. Not only do they not want to waste resources on this, the banks probably don’t have access to the resources necessary. Especially credit unions.
they want to not pay your money back when an AI agent empties your account
same as when a family member empties your account & they accuse you of being complicit when you try to get your money back & say they will prosecute YOU for fraud for having authorized it (even though you DIDN'T)
i guess you can tell i've known people this has happened to
Shouldn't Meta be very irritated by this?
@futurebird culturally, banks are not into technology.
@futurebird Look at the money they're lending, and how the debt is structured for data centers and computing companies versus the aggregate of the credit market. Do they have it easy, about average, or are they paying more and getting less, like any other high risk borrower?
Like an energy company that pours money into new wells and more efficient solar panels, it's probably a mix of projected risk and present gain.
Aside from general regulation, banks need insurance to cover risks. Whatever the preferred stance of individual enterprises, I can see insurance companies refusing cover if the door is left open.
@futurebird
it only seems wild if you believe an agent with your bank information is a "service".
It is not a service. It is yet another way that tech companies steal from people. It is robbery.
That it is a service is the pretext it operates under.
@futurebird A false pretext. Every time a so-called "AI agent" uses bank information to make a purchase, the company who created the agent commits wire fraud.
Install Fedizen
Get the full app experience
@futurebird "We've brought your banking into the future, now you can access your account with our agentic AI, just say things to the chat window like "put all my money in crypto" or "buy apple giftcards with my current balance" and instantly it's done! No more calling a customer service agent, just 24 hr service."