AI doing its own thing with online payments?

aurigacap

Autonomous action by AI has already spread online payments, as well as AI developers. A cautionary tale.

Last night I tried to buy some unexciting car parts online from a well known (ie. legitimate) Audi supplier. I did so using a link supplied by them with a pre completed on line invoice. I tried to pay with Barclaycard: payment rejected by the supplier twice, not apparently because my card details were incorrect....just a long error number without human explanation. Retried with bank debit card… same result. Funny, I thought, so I checked both accounts online. Two Barclaycard payments for the amount were pending and one bank payment……but not to the seller in question, to an entirely different and apparently unrelated payee. I established this by checking in my account with the supplier: no orders invoiced, and by exploring whether they used an alias for payments: not this one.

Naturally, I reported all this to Barclaycard and my bank immediately, thinking that it was better to block the payments pre-emptively while pending, rather than wait until the money was paid. Besides I reckoned it would be to my advantage to secure a refund, in the event, if my early action was registered on screen. In fact, both advised me to wait until the payments were made(!) I did all this over 2 hours on the phone, last night and this morning because there was no obvious route for reporting a disputed pending payment online or on my banking app. (That's another story).

Last night, I also emailed the legitimate supplier with the facts and suggested that their payments systems may have been hacked. When I phoned this morning they revealed the AI angle. This was not the first time, though the pending payee was new. Their 'secure' payments systems provider confirmed that their AI was both blocking some legitimate payments to them, then redirecting them to randomly selected payees generated by AI. They assured me that the payments would remain pending and would not be drawn from my accounts.

We shall see. I'm not naming names yet, because the full facts have yet to emerge. But it seems that we will have to add autonomous AI fraud to the list. Who will be liable? I assume those who deploy it. All this for mud flaps.

eskbanker

I think you're right to use a question mark in the thread title and would take the AI references with a pinch of salt…

On the subject of (in)ability to dispute payments while pending, this has come up before and is simply a factor in how payment protocols work - once a card provider has authorised a transaction, they are irrevocably guaranteeing payment, so don't have the option of reversing that prior to settlement, and any corrective action has to wait until the payment is completed.

born_again

Pending payments can not be stopped, as using debit or credit card is a guaranteed form of payment. Wrong company name on a payment is not a dispute right.

As a aside, ever get a declined payment, never try again. Always check bank to see if payment is showing as a pending amount. Some retailers systems time out & show as declined, but bank has already approved payment.

Jemma01

Rather odd putting autonomous AI so close to payments! They're not ready for that.

MyRealNameToo

There may also be a simple case of chinese whispers, the person who answers the phone to a customer doesnt work in the IT department, it wont even have been the IT department that wrote the notice to give to the customer facing staff but a whole chain of people… this is then compounded by the fact you are talking about crossing two companies…. sure if you are a multinational then you may get some IT to IT comms but if the people you are buying from are a more run of the mill company then they too get communicated to by a customer facing member of staff not a technical expert.

Long chains of people rewriting the message means what the OP has been told may have little resemblence to what has actually occured.

I'd echo that true AI being involved in the core payment flow sounds highly improbable but maybe there has been a basic coding error and maybe AI was involved in creating the code… thats about as close to it being AIs fault that I can think is realistic and still its really the fault of the testers who didnt spot the error.