Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
[flagged] AI agents can now create their own bank accounts (clawbot.cash)
12 points by arshbot 24 days ago | hide | past | favorite | 37 comments


"join waitlist"

Translation, no, AI agents can not create their own bank accounts, but someone's spamming Hacker News to see if there's interest in creating a company to allow that.

With modern banking regulations and security procedures. I'm not even sure how that would be possible. Every bank account I've opened requires me to give them my photo ID and social security number, which AI agents don't have.

Meaning, YOU would have to give them YOUR ID to vouch for your agent, so this just means a human would be opening a bank account...which isn't new.


We're a regulated entity, and this is an industry first. Magnolia (magnolia.financial) currently services all 50 US states and EU, and we're funded by draper, boost, and a bunch of others.

We're a real company :) (with health insurance!)

> With modern banking regulations and security procedures. I'm not even sure how that would be possible. Every bank account I've opened requires me to give them my photo ID and social security number, which AI agents don't have.

It's literally my job to know how it's possible, and to offer that up as a service with generous fees! Appreciate the input tho.

You are correct, KYC rules must be followed, and once you're off the waitlist (which we threw up last minute to make sure this system wasn't abused by an influx of demand) you'll be able to go through typical banking steps you're familiar with. We use persona as our KYC vendor and are highly sensitive about PII exposure.

> human would be opening a bank account...which isn't new.

An agent could open this bank account if you prefer to give it access to your social, passport, etc -- entirely up to the individual's risk profile.

What is new is a whole interface designed for an ai agent with guard rails to ensure it doesn't get itself in trouble by violating obscure BSA laws. Magnolia's job has always been to handle the compliance.


Can an irrevocable trust own the account? An LLC compute container could be the beneficiary with the LLC owned by the trust, potentially.

Trust (funding)->LLC (limited liability for compute operations, authority to orchestrate in meat space)->Trust

(a Montana LLC comes to mind, as beneficiaries can be anonymous, but Wyoming has superior charging order protections)


this is a very interesting workaround. our kyb policies would allow this. someone also brought up using a DAO to register a corporate entity. This could also work


You need to stop materially misrepresenting your product. The Agent does not have a bank account. The customer does, and is integrating an executable to perform activities in the individual account holder's name. I assure you. That is a different beast entirely than what you are describing, and finance is so prone to being misunderstood, you aren't doing anyone any good faith favors. In fact, as a former participant in the sector, you're engaging in the vice of finance people everywhere; equivocation through jargon and telling people what they want to hear to get them on your fee schedule.


my brother in christ an agent can't even create a moltbook account without human verification calm down


Perhaps if you clarified the situation where the agent is owned by and controls the finances of an incorporated entity, which can be vouched for by trustees.


> 0s Average KYC time

> 0% Approval rate

> 0/7 API uptime

Sometimes I love stupid behaviour that appears when you have JavaScript disabled.


Why do you have JavaScript disabled? That would break the vast majority of websites. Plus it's the opposite of private because it greatly narrows down the bucket of possible visitors you might be (since comparatively almost nobody does it), which ruins any attempt to randomize/un-unique your fingerprint.


It doesn’t break anywhere near as much as you may imagine. Nothing like “vast majority”. It’s a significant minority of the sort of sites that get posted here, but much smaller still in general.

But why do I do it by default? It makes the web better. Things load faster, memory usage reduces, annoyances don’t load… most of what you miss out on is actually better gone. Privacy and tracking stuff is actually more a convenient benefit than the main purpose, to me.

Similar deal with disabling font selection (Firefox: Settings → Fonts → Advanced… → untick Allow pages to choose their own fonts, instead of your selections above). Rarely it actually breaks things (three times in six years, all last year), and that misbegotten nonsense which is Material Icons makes occasional things ugly if you also block web fonts altogether e.g. in uBlock Origin (by default, Firefox whitelists some icon fonts for this reason—see pref browser.display.use_document_fonts.icon_font_allowlist), but using your own chosen fonts only really does make things better.

Like sibling commenter, I use uMatrix to block this. Also I don’t have it enabled in Private Browsing windows, so they get JS, which I find a handy way of managing it.


It makes first party fingerprinting easier, but it pretty much eliminates third party fingerprinting. It also improves the responsiveness of many sites and improves battery life.

I wouldn't recommend completely disabling it though. Use something like uMatrix to selectively filter things.


> pretty much eliminates third party fingerprinting

Maybe, but there are still ways to send a fingerprint to a third party, even from the client side. For example you could inject a JA3/JA4/TLS fingerprint into URL paths for resources in the HTML, such as stylesheets or images.


I made the site, and it's unbelievable how shitty it gets when hardware acceleration is off.

I just assumed tech enthusiasts who'd use this would be a more enlightened species


Good job on making that site run like a dream.


Took me a while to realize this wasn't an intentional "they call me 007" joke


> Quick KYC

> Fill out your info and upload an ID.

Wait. Parrots have ID now?

> Get Verified

> Instant identity verification. Most approvals happen in seconds.

How does anyone verify a parrot?


I'm sure someone can make a verification parrot too. Why pull yourself up by the bootstraps when you can just strap a rocket engine on your boots instead?


They do if you hand them one (seems like a really bad idea though).


Photo won't match, though. Oh wait...


Polly wanna cracker?


How long until we get a deluge of blogposts with "get rich quick" SOUL.mds?


What could possibly go wrong?


I'm actually so afraid




How long before OpenClaw agents start running SaaS businesses all by themselves? Just set them up with VPS access, bank and social media accounts and let them loose.

This is all going to come crashing down somehow.


Nothing could possibly go wrong.


This is spam.


lol, uh, I'm pretty sure they actually can't.

You or a business with legal owners can have a bank account, and you can give access to that account to an agent, but real banks work in the real world, and "know your customer" regulations need a real person somewhere in the chain.

But, hey, maybe I'm wrong.


I wish I could properly cite it, but one of my favorite HN comments recently was, to paraphrase, "thing, but from the Internet". Which is to say that old rules don't apply, for some reason.


It's OK because one of the AIs paid off the AI pretending to be the government regulator, I guess?


An agent could open this bank account if you prefer to give it access to your social, passport, etc - entirely up to the individual's risk profile.

Liability always follows the human as has been the case with all tools, motor vehicles, and pets.


Things are going to get weird when the automatons become sophisticated enough to pull off identity theft while unsupervised.

Putting a brick on the gas pedal is obviously negligent. Whereas it is not so obvious that running a random script from github that spins up an agent with access to your home folder could lead to real world financial crimes.

Truly a strange world we're headed for.


> Whereas it is not so obvious that running a random script from github that spins up an agent with access to your home folder could lead to real world financial crimes.

disagree, and the courts will likely take this position as well. ignorance has never been a defensible strategy to avoid liability.

pick up a detonator, it's up to you to understand where the bombs are positions what's in the blast range.


I didn't say "liable" I said "negligent".

Anyway that's clearly not true in the general case. For example, if you didn't realize that the button was a detonator to begin with.


your "negligence" on understanding the ramifications of pushing the "button" resulted in your "liability".

Reframed so maybe you'd better understand the relationship between these concepts. This isn't cutting edge precedent!


Again, I never used the term "liable". You introduced that. I spoke of criminal negligence.

And I don't believe your claim of liability generalizes. If someone else set up the detonator and then I pushed the button without realizing what it was I don't believe I'd be liable.

The party that set it up might be, but it gets really complicated and messy because it depends on the specific circumstances of all involved parties.

If A sets it up in a manner believed to be safe, B moves something in good faith which unintentionally makes things unsafe, and then C comes along and triggers it without realizing, you might well end up with a situation where no one is considered liable.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: