Is it safe to connect my UK bank to ChatGPT or Claude?

Connecting a UK bank to ChatGPT or Claude is safe when it goes through open banking: you approve the link in your bank's own app, the assistant never sees your password, and you can end the connection whenever you like. That is how Era does it, and it is a very different thing from handing an assistant your login.

What makes a connection safe or unsafe?

The risky way is to give an assistant your banking username and password, or to let it operate a browser while you are signed in. Banks cannot tell that apart from someone breaking into your account, which is why some are tightening their terms around it.

The safe way keeps your credentials out of the picture entirely. You sign in to your bank yourself, in its own app, and approve a limited, named connection. The assistant then asks Era for information, and Era asks the bank only for what you allowed.

How does open banking protect you in the UK?

Open banking is a UK system, regulated by the Financial Conduct Authority, in which you tell your bank that a particular provider may read your account data. You give that consent inside your bank's app, using the same sign-in and biometric checks you use every day, and you reconfirm it with the provider every 90 days.

Barclays, Lloyds, NatWest, HSBC, Monzo, Starling and Revolut all support it. Nothing in it lets an assistant move money on its own, and you can withdraw your consent from your bank's app or from Era at any moment.

Who is liable if something goes wrong?

Because the link to your bank is read-only and you approved it in your bank's own app, an assistant cannot spend your money. Anything you do after a suggestion remains your decision in your own banking app. Money movement through Era is in private beta, off by default for every key, and every transfer needs your approval.

The FCA has said it plans to work with industry on a trusted framework for agentic finance, including clear approaches to agent identity, delegated authority and accountability, so the question of who is responsible is being settled in the open and in your favour.

What can the assistant see, and what can it never see?

Your assistant sees balances, transactions with merchant names and categories, spending analysis, recurring payments and the financial context you have chosen to share. It never sees your bank password, your full account numbers or anyone else's data, and it receives structured answers rather than raw database rows.

Every request is recorded in an activity log you can open at any time, showing what happened, when, and which assistant did it.

Can I disconnect an assistant, and does my data train a model?

You can disconnect any assistant in a couple of taps, and the others carry on unaffected. Authorising Claude does not authorise ChatGPT, and removing one does not remove the rest. You can also end the bank link itself in Era or in your bank's app.

Your financial data is not sold, not used for advertising and not used to train AI models, and that applies to Era and to the providers that process requests for it.

What do I get once it is connected?

You get one hub that every assistant uses, so Claude, ChatGPT and Gemini all see the same accounts and share the same memory. You can share a space with a partner, have statements and payslips read and filed, and watch your financial health figures update alongside the analysis behind them.

Era connects accounts across the US, the UK, Canada, France, Germany, Spain, and 40+ countries in all. For the full privacy model, read how Era protects your financial data, then connect Claude to your bank account or connect ChatGPT. Wondering whether your bank has its own connector? See does my UK bank have an MCP server.