Banks find AML "ineffective", propose access to social media

Even if a piece of information isn't informative on an individual basis, it might provide improve the performance of a model that combines it with other inputs.

You don't need to posting about doing crimes for your FB profile to be useful for risk assessment. Your connections or other posts (e.g. about how some emergency means you're suddenly short of cash) could contribute to an elevated score.

The proposal doesn't go into much detail about how they imagine it will work, but it sounds like they're just saying that a statistical model (based on a larger set of data) would be more useful than a set of rules based on a tiny set of inputs. This seems entirely reasonable.

Here is the relevant paragraph:

  Input data for MSA platforms should incorporate not only transactional data, customer static data and internal reference lists, but also other dynamic behavioural customer information where proportionate to the risk (e.g. device ID, IP addresses). Using an RBA, input data may also include data from reputable external, publicly available sources, including information on company structures, Ultimate Beneficial Owners (UBO), and watch lists, as well as complementary sources such as market  data and verified customer social media accounts. Finally, FIs should establish robust data governance and quality control frameworks.

The relevancy of advocacy groups is at an all time low because of just how little the professionally informed care these days. Tech was a "community" when the grifters slid in wanted something from the rest of us, and YOLO fuck you I got mine backstabbery when anyone suggested maybe not selling out our privacy quite so hard.

And it infects the rest of us. When our friends and colleagues have jobs swinging pickaxes at our societal expectations and standards of privacy, why do we pretend they're morally neutral in the whole thing?

When I moved to the UK, I was advised to open a bank account with either Monzo or Starling, which are the two online-only startup banks in the UK. (They're actual banks, not resellers backed by other banks, or money transfer apps like CashApp.) This is because their identity checks involved scanning a single identity document and sending a video of myself saying that I wanted to open a bank account. All the traditional 'High Street' banks wanted all sorts of documentation - Council Tax bills, infamously - that I simply didn't have, because I'd just arrived in the country.

I know why all this is necessary. London remains the financial capital of the world, and launders a lot of dirty money. Monzo in particular has strict limits on cash withdrawals and deposits, because their accounts are often used in the illicit drug trade.

I have no objection to AML and KYC regulations, but it's absolutely crucial that legitimate people who don't fit the usual patterns - people like recent immigrants - have access to the banking system.

The social media aspect of TFA is attention-grabbing, but I think it's a red herring. The more concerning aspect is the suggestion of AI. To what extent will those algorithms be used, and to what supervision will they be subject?

In the 1980s, my father was an 'Automation Engineer' in a bank. He built an 'artificial intelligence system' for assessing home loan approvals. It was one of those fill-out-a-form-and-we'll-tell-you-how-much-you-can-borrow things that are all over the Internet now - entirely deterministic and transparent.

That's how banking needs to work, at least for individuals. Determinism engenders trust.

Modern AIs, on the other hand, are for algorithms that are too complex to be explained. If you've got those controlling people's access to finance - not just loans, but money they've deposited in their own bank account - we have a serious problem.

'Cash is King' is not a solution. Before my Dad worked at the bank, he worked in the payroll department at a soap company. He had to drive to the factory with the paypackets - one for each employee, containing their wages in cash - each week, and hand them out. That's utterly ridiculous.

> Your connections or other posts (e.g. about how some emergency means you're suddenly short of cash) could contribute to an elevated score.

This sounds more like a way to send desperate people into a death spiral by raising interest rates or denying credit availability right at the worst possible moment than any kind of system that would prevent crime.

> The proposal doesn't go into much detail about how they imagine it will work, but it sounds like they're just saying that a statistical model (based on a larger set of data) would be more useful than a set of rules based on a tiny set of inputs. This seems entirely reasonable.

It's implying they want to invade everyone's privacy to set up a hallucinatory black box that destroys lives at random under the guise of... I'm not even sure what this is supposed to do. It's not going to yield anything that would be useful in a prosecution so it's more like opaque government-facilitated denial of service without due process. Far from reasonable, it's total crap.

> Monzo in particular has strict limits on cash withdrawals and deposits, because their accounts are often used in the illicit drug trade.

ATMs often have fees that the bank will pay for. For example, most ATMs have an upsell at the end to "check your balance" that your bank will pay for. On the other hand, card payments generate a small amount of revenue for the bank. Most payments in the UK are card payments or bank transfers.

I'll have to say you're mistaken. Monzo doesn't have limits on cash withdrawal as an AML control.