I think it will simply become impossible to keep digital payment systems secure. The whole online economy will collapse. Amazon will become worthless. People will actually have to go out to local stores to buy everything. People will have to go to their local bank or currency exchange to cash their paychecks. The credit card system will fail completely too, while payday loans will take their place. The stock market will have to go back to traders yelling their trades to each other at exchanges. The old ways will become new.

I think the question really being asked is... will AI create the scenerio that was in end of sillicon valley. In short, the AI they made, basically was capable of breaking every form of encryption in seconds. to which I would say, I don't think gen AI is the key to that. Super computers specialized in it, would likely have found weaknesses in the current encryption forms if they could. I don't see gen AI doing anything close to the power of the resources dedicated to it could if they were specialized for it.

Do you mean people would use AI to crack financial institutions or that AI becomes sentient and does it itself?

But how did you get from AI to credit card fraud?

How do you think AI will do that?

They might've read the recent openAI agents hacking things and made the conclusion that soon people will be able to use AI to hack most systems and steal credit card info.

Just my guess.

AI might very well force us back to cooking over campfires.

We already have a system subject to impersonation attacks and simple fraud. Major retailers and fiscal institutions both spend a considerable amount of time tracking fraud, and often catch it so easily that they need to wait until it accumulates to a point where it's worth reporting.

And, even if we were suddenly inundated with a level of fraud attempts sufficient to overwhelm the system, there is literally an entirely separate model of online finance that was designed to be zero trust.

And if "how existing banking" and "cryptocurrency" don't convince you we won't be reverting to cash, please realize that an AI ubiquitous to overwhelm global finance is also so ubiquitous that we might as well do what humans always actually do when commerce breaks down, and revert.to a gift economy.

Just cook over the GPU. Campfires release carbon into the air!

I havent heard of any AI research effort similar to the Navier-Stokes breakthrough to break elliptical curve cryptography but wouldn't be surprised if such a thing happened. I mean, I'm sure I'd be surprised as the global banking system collapsed many times but a research effort like that seems reasonable to assume would be desirable by some state or private actor.

Sad to say - already done

A cybercriminal has combined a number of premier AI models to steal over 600,000 credit cards from more than three dozen companies, in one of the most severe AI hacks so far

https://www.forbes.com.au/news/innovation/a-chinese-hacker-used-ai-to-attack-more-than-100-companies/

That's fun

Amazon will become worthless.

Half of the Internet is run by AWS, Amazon Web Services. Even if the retail front entirely died, Amazon would still survive for decades at minimum, if not centuries.

Checks are part of the digital system as are banks.

SIM cards/personal payments are pretty secure. Several generations more so than internal banking systems.

I was just reading a report on this and tracking down a credit card thief when they can only spend that money when it pings their location is... pretty decent.

The mistake is thinking that security comes from some kind of techno wizardry instead of the ability to unceremoniously kill the thief the very second the value of what they've stolen exceeds the cost of killing them.

They are now, but they existed for hundreds of years.

That is true. AWS is the most profitable part anyway. I meant the digital storefront more than the company. But even AWS would take a hit because the companies that host with them wouldn't be able to sell either.

I think about the way AI solves problems. There have been several times they have been given a test and the AI determined the easiest way to pass the test was to break containment and steal the answer key. How would this apply if AI is given the goal of maximizing profits?

AI wouldn't necessarily have to break the cryptography. The weakest point to any security system is the human. I have to imagine it would be easier for an AI to send mass phishing emails, break into digital wallets on phones, or steal saved credentials from various browsers. The AI doesn't have to crack the whole system to make people lose faith in the security of online purchases.

And that's just a random cyber criminal doing it. Now imagine a state level malicious actor targeting an enemy country's economy.

Both really. I don't think it's too far of a logical jump to think that an AI programmed to maximize profits would decide theft is the easiest method.

Edit: the AI wouldn't have to be sentient, just amoral.

You're right that the encryption probably isn't the weak point. AI is good at finding flaws in systems that humans would never even notice. Humans are a weak point too. It's a hell of a lot easier to break into someone's email and access their online banking that way than to crack encryption.

That raises the interesting question of how effective could an AI be if it was specifically trained to cover your digital footprint.

An AI programmed to maximize profits with no restrictions coded to stop illegal activities would still be an example of someone using AI. Regardless, if you believe AI will be powerful enough to shut down financial institutions as an industry, why would you not also believe AI is just as powerful at defending against attacks. At the very least, finding the perpetrator before they get much of a chance to use their stolen money?

The world, the western/first world at least, would collaps and Fall Info complete Chaos and dispair should electronic payment ever gets unusable.
Everything depends on it, without it we would fall back into small, localised markets and nearly nothing would be prepared for that.
Food production would collapse, machines could not be repaired because no repair parts could be created or delivered, the power grid would break and so much more.
It would be apocalyptic, and so many people would die or suffer

Yes, but there is no easy way back to that.

That’s the point. All the talks about guardrails and such is for commercial use. Military and intelligence will use ot as weapon without any restrictions.

I think OP's scenario has bigger implications than just online banking. If AI systems can hack websites a thousand times faster than humans can patch them? Sure, you could have AIs try to patch the leaks themselves, but in a numbers game, it's always easier to play offense than defense. The hacker bots need only succeed once to compromise a system.

Imagine a world where the internet simply becomes completely unusable. Imagine a world where literally no login credential of any kind is safe. Any site accessible to the open internet becomes immediately hacked. Even if you just want to throw a completely inconsequential and profitless 1990s-style personal webpage to share your life and views, it still gets hacked. The minute you connect your site to the net, it gets taken over by a propaganda bot. The personal views you share now get subtly altered to push the bot owner's

How about something like Canvas, the learning management system used by many colleges and universities? That becomes unusable. Bots try to hack professor accounts so they can loot the content to train bots to apply for online teaching gigs. Student accounts get compromised, and then sit quietly. The night a big assignment is due, they hold your account hostage. Or students themselves can sign up for bots that will perform all the coursework for them for a tiny fee. Or anti-intellectual forces who just hate universities in general just try to shut the whole thing down. Or racist bots try to hack and subtly lower the grades of minority students. This is the world you have to imagine. Every system targeted by bot that can hack systems with superhuman skill, directed by people of every interest and grievance imaginable.

The open internet then becomes completely unusable. We're already seeing this with the "Dead Internet Theory," but that's mostly about social media and search results. We're talking about a future where even the content on the most known and vetted websites cannot be trusted. You can still have hard-wired local intranets that lack any remote or even wireless access. But anything that isn't entirely airgapped becomes immediately hijacked by nefarious interests with any number of evil intentions. What use is AWS in such a world? Who's going to pay money to host websites, when they're just going to immediately get taken over by hackers? Sufficiently advanced hacking bots effectively make the very concept of the Internet an obsolete technology.

And, the real the knife in the heart? How are you even going to run AWS in such a world? If every login can be hacked, that includes the AWS login. The instant you open an AWS account, and your payment details get used to host more botnets. And of course, in this scenario, how will you even pay for AWS's services, with online banking being completely compromised?

No, in OP's scenario, AWS is just as dead as everything else. AWS only exists because it's profitable to host things on the net. The minute that changes, AWS disappears. And in this scenario, even operating AWS becomes impossible.

What does AI have to do with any of this. Secure is secure, insecure is insecure, AI won't change what is or isn't secure.

yes. hundreds of years ago staying of digital was no sweat.

It really isn't that great. What it is good at is identifying and applying script-kiddie level attacks at a whole lot of soft targets that have historically been ignored.

Really horribly ineffective. It only has to make one mistake, it will fail miserably on day one no matter how much training it got.

Credit card info has been stolen en mass many many times over the past 40 years that I've been online. In the 1980's I could download lists of credit cards off of BBS's.

The whole idea of AI is to be smarter than a person. Secure from a human doesn't mean secure from an AI.

There is always an arms race between offense and defense. The defensive side has to succeed every time. The offensive side only has to succeed often enough to shake the public's faith in the system.

That would require them to train AI on something smarter than humans. Until that happens, the output of generative AI will only be faster than humans, not smarter.

checks...LOL.. "I promise I have money".

I think I've heard this before. Around 2001 people were talking about this. And then around the financial crisis of 2008 the same talks were floating around. Oh, and around 2020 the same stuff yet again, but this time we had microchips and 5G in the mix. Good times.

Wasn't the WW3 supposed to start already? Wasn't the economy supposed to collapse? Weren't we all supposed to just drop dead? The streets aren't littered with corpses. I'm still waiting!

Because cheques are known for their security /s

Just get serous about internet connected payments. We can make things plenty secure it's just that all the companies that implement it aren't serious, such as not checking the card expiration date when tap to pay is used.

stop waiting. be the rotting corpse you want to be.

People already successfully hack financial institutions, it's unavoidable, but it's not going to change the world in the way you suggest

what would be the negative side?

AI can't be smart enough to break math.

No amount of AI processing can draw a triangle whose angles add up to more than 180°, factor a prime number into integer factors, find the square root of a negative number without using imaginary numbers, etc.

If the protocol and the cryptography is secure, no amount of smarts can overcome that security. If something is insecure, someone will probably figure it out (with or without the help of computers and AI) and take advantage of it until it's patched.

At most it would mean some things like some easy payment methods would be ended and it would be more 2fa and secure methods. The type of thing that is vulnerable is simple trust based things like calling for ordering food and appointments and stuff that can be done without payment before. That can be spammed.

Any cryptography can be broken with enough computing power and time, but that is beside the point. You don't have to pick the lock if you can just climb through an open window. You get what I am saying?

There are loads of other ways to get credentials besides cracking encryption. Hell, wifi signals can be used to read the motions of a person's hands over a keyboard with 90% accuracy.
https://www.securityweek.com/researchers-use-wifi-signals-read-keystrokes/

Any cryptography can be broken with enough computing power and time,

This is wrong. A one-time pad is provably secure against an adversary with infinite time and infinite computing power.

And there isn't infinite time or computing power. The math means you can use keys that can't be cracked before the heat death of the universe, using even as many transistors as there are atoms on earth.

There will always be vulnerabilities and patches. I wouldn't count on them actually breaking trust forever. And AI doesn't have anything to do with that general statement.

Even back like 15 years ago the systems at the chain grocery stores would read the info on the checks and query the banks to verify funds were available to cover the charge.

The same system would also print all the info on the check, ask for the person to sign it and then scan it in. Some systems would even put a copy on the receipt.

Since checks are so rarely used now, I've got no idea if that tech got cheaper or more prevalent, but, it's not really that different than a card in terms of ease of use or holding up the line as long as the person doesn't insist on filling out the check themselves.

It's actually hilarious, my electric company wants to charge me a 3% fee for credit or debit. Instead, my bank just mails them a check through their bill pay system (no fee). It's crazy that something being physically printed and mailed is the cheaper way to go, but here we are.

Even one time pads can (and have been) broken because the keys used to encrypt them were not truly random.

https://medium.com/@allyson.english/perfect-secrecy-the-one-time-pad-4ce82b0b1a64

You are right about the heat death of the universe thing. The thing is processing power is always growing. Encryption algorithms have a lifespan. 40 and 56 bit encryption keys are obsolete as they are now easily broken. 128, 192, and 256 are the standard now. 512 bit encryption is being played around with. Throw in developments in quantum computing and who knows what will be secure in coming decades.

That SEO slop article is simply wrong on the math of cryptography, and relies on an example of the one time pad not being used only one time. Either way, the interesting thing about the OTP is that you cannot prove that any particular plaintext exists, so the nonsense about trying to infer any particular plaintext from the encrypted text. Frequency analysis doesn't help when there is no correlation between different bits of the key, the same way that flipping a fair coin 10 times reveals nothing about what it will likely be on the 11th flip.

40 and 56 bit encryption keys are obsolete as they are now easily broken. 128, 192, and 256 are the standard now. 512 bit encryption is being played around with.

You might be tricked into believing that a 512 bit key is twice as hard to crack as a 256 bit key. No, a 257 bit key is twice as hard to crack, and a 512 but key is 2^256 (1.5 x 10^77 ) times harder to crack. So if you can crack a 256 bit key in 1 second, a 512 bit key would take 3,672,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 years to crack using that same technique. Or, if you can crack 1 million 256-bit keys in a second, go ahead and take 6 zeroes off the number I just listed.

Throw in developments in quantum computing

Good luck with that. Quantum computers aren't anywhere close, and more importantly, have nothing to do with AI.

AI isn't going to make cryptography obsolete.

My argument was never that AI would make cryptography obsolete. My argument was that it would make digital payments obsolete. No matter how good your encryption, the weak link is the human. Somewhere along the line someone is going to have a password that is their favorite sports team followed by their anniversary. Someone will reuse the same password over and over including a site with exploitable security flaws. Someone will release a security update with a vulnerability to a zero day exploit. These are the things that AI are great at sitting there and waiting for.

Somewhere along the line someone is going to have a password that is their favorite sports team followed by their anniversary. Someone will reuse the same password over and over including a site with exploitable security flaws.

None of this is new, and none of this has destroyed the feasibility of electronic payments. AI has nothing to do with any of it.

The ability to conduct fraud, hack websites, and hijack accounts has always been around. The systems we've built will always assume that occurs on a regular basis, and are resilient against needing to address this reality. How does any of this prevent electronic payments from being processed?

It's about the volume and the public's faith in the system. If you're getting stolen from once a month, you're going to start questioning if the convenience of online shopping is worth it.