OpenAI has added a voice feature to ChatGPT Finance, its money tool. Instead of typing, users can now speak their questions about income, spending, transactions and savings — and interrupt the answer partway through to ask something else.
The interruption detail is the one that tells you what they are building. Being able to cut in mid-sentence is how people actually talk, and it is a deliberate attempt to make the tool feel like asking a person rather than querying a database.
The condition that has not changed
None of it works until you connect your own financial accounts to ChatGPT Finance.
That requirement is easy to skim past, and it is the whole story. The voice layer changes how you ask. It does not change what the tool must be given in order to answer: continuous access to your transaction history.
Where it works
The voice feature is initially available only in the United States. ChatGPT Finance itself began as a US pilot on 15 May, restricted at first to Pro subscribers, and was later opened to Free and Go subscribers in selected countries that have not been named.
Earlier features include weekly updates analysing your financial situation, and current information on shares in companies you choose to follow.
What the announcement does not say
This is worth stating plainly, because the gap is large. The report covers what the feature does and where it is available. It says nothing about how the system processes financial data, exactly what it can see once connected, how long it retains any of it, what security protects it, or what consent is actually being given at the moment you link an account.
Those are not pedantic questions for a product whose function is to read your bank statements. They are the product.
What it would take here
Nothing in this is available in Bangladesh, and the obstacle is structural rather than a matter of waiting your turn. A tool like this needs a standardised way for a third party to read your account with your permission — the open banking rails that exist in the US, UK and EU.
Without that, there is no safe mechanism for an app to see your transactions. There is only the unsafe one, which is handing over your banking credentials to something that logs in as you. Anyone tempted to replicate this locally with screenshots of their statements should note that they would be uploading their full financial history to a third party with no agreement covering it.
The honest summary is that the feature is well built and the question it raises is unanswered. Asking aloud is pleasant. Knowing what is on the other end of the microphone matters more.
The questions worth asking before connecting anything
If this does reach here, or if you use it while travelling, four questions decide whether it is safe to link an account, and none of them requires technical knowledge.
Is the connection read-only, or can the tool move money? Can you revoke access from your bank’s side rather than only from the app? Is your transaction history retained after you disconnect, and for how long? And is any of it used to train models?
A service that answers all four clearly is making a commitment you can hold it to. One that answers none of them is asking for your financial history on trust, which is a great deal to extend to a product announcement.
