Your bank holds the most detailed record of your life. 13 of 19 won't say if it trains AI on it.
Every story about AI training is about a chatbot. Meanwhile the institution that knows every payee, every balance and every transaction you've ever made isn't asked. So we read 19 privacy notices, 26 documents in total, looking for one thing: whether your data trains a model. 13 of them never address it. Not refused: unaddressed, in documents that run to tens of thousands of words about your data.
One of them asks the question itself, then answers around it
Starling Bank devotes a section of its privacy notice to artificial intelligence. The heading is a question, "Do you use artificial intelligence when processing my data?", and the section answers it without ever saying whether your data trains a model. This is the clearest example on the site of a policy raising the subject and leaving it exactly where it started.
"No external company will keep access to any Starling data, and no external AI models are trained using Starling data."
One says it plainly
PayPal is the only one here that states it outright. What surrounds the sentence is the finding: the same notice carries opt-outs for personalised shopping, for advertising, for disclosures to partners and merchants, and none for this.
"We may use Personal Information to train our artificial intelligence (AI) models that power our Services and help us deliver more secure, efficient, and personalized services."
And one incumbent high-street bank breaks the pattern. Lloyds Banking Group lists training systems that learn from your data among its uses, with legitimate interest as the legal basis, which is also what gives you a right to object. That it's the only incumbent to do so is why the silence elsewhere gets named, not assumed.
The ones that say enough to grade
- FCommonwealth BankCommonwealth Bank says it uses customer information — transaction records among it — to train artificial intelligence models. Australian customers are offered no way out of it.
- FPayPalPayPal states plainly that it uses your personal information to train its AI models. It offers opt-outs for advertising and personalised shopping, and none for that.
- FRobinhoodRobinhood says your inputs and conversations with its chatbots and AI features may be used to train its AI. It names no way to decline, only consent "where required under applicable law".
- DLloyds Banking GroupLloyds is the first incumbent bank here that answers the question: it trains systems that learn from your data, and says so in its own words.
- DMonzoMonzo does train AI on customer data — but only where it cannot anonymise it first, and it bars its AI suppliers from training on you at all.
- DRevolutRevolut trains internal models on customer data, including credit decisions made by AI with no human input. You can object — but it warns that objecting may cost you the account.
The ones that never address it
UNCLEAR is a finding about the document, never an accusation about the company. Each of these was read and searched for training, machine-learning and AI language. Each verdict records which documents were read and which terms were looked for, so you can check the search instead of taking our word for it. A bank may well not train on you and simply never mention it. What none of them does is say so.
How to check any of this
Every quotation on this page is word for word from a document we archived, and every verdict links to the copy it was taken from with the date it was captured. The most recent of these notices was re-read on 7 October 2026. If a bank changes its wording, the change appears in the record with both versions side by side. If we have a clause wrong, send us the paragraph and the correction gets published beside it.
The method, including what UNCLEAR does and doesn't claim, is set out in full on the methodology page. The same silence read across every category, not just this one, is on patterns.