By J. H. Irwin
Author | Content Creator | Technology Strategist
A couple of days ago, ChatGPT asked whether I wanted to connect my financial accounts so it could help me manage my finances…
I was caught off guard. My answer came quickly and without hesitation: No.
Not now. Not any time soon.
I am interested in what artificial intelligence can contribute to our lives. I can see the value in a tool that explains a complicated subject, organizes information, or helps us think through a problem. But an invitation to connect my financial accounts felt considerably more consequential.
My money represents years of work, obligations I have to meet, and whatever security I have managed to build. Before I allow another service into that part of my life, I want to understand precisely what I am opening.
And I want more than a reassuring invitation to get started.
That requires an important distinction. Connecting a financial account does not automatically authorize an AI service to transfer money. A connection limited to reading balances and transactions is different from one that can initiate payments or trades. What a service can do depends on the integration and the permissions granted.
OpenAI’s documentation for ChatGPT Work describes connected access as bounded by account permissions, approved scopes, and available actions. It also explains that some authorized changes can proceed without a new prompt when the configured policy permits them. Those general controls do not establish what the particular financial invitation I saw would have authorized.
I did not proceed far enough to examine its permissions. I cannot claim that accepting it would have given ChatGPT the ability to move my funds.
But I can say this: I should be able to understand the boundary before I am asked to cross it.
The difference between viewing my money and moving my money needs to be unmistakable. So does the difference between access I grant today and any additional access a company might ask me to grant later.
A helpful conversation should never leave a person guessing about either.
Even access limited to reading financial information deserves scrutiny. Our transactions can reveal medical expenses, charitable giving, travel, family responsibilities, and periods of financial hardship. A record of what we spend can become a remarkably intimate account of how we live.
Before sharing that record, I want to know who receives it, why they need it, how long they retain it, and what happens to information already collected when I disconnect.
Those are reasonable questions to ask any company handling sensitive information.
The potential benefits are real. An AI assistant could help someone understand a budget, notice recurring expenses, or prepare better questions for a financial professional. For a person overwhelmed by paperwork or unfamiliar terminology, clearer explanations can make a meaningful difference.
I would welcome technology that gives people a better understanding of their finances. That understanding should include the technology itself.
There is a danger in assuming that because a system explains something fluently, its judgment is dependable. In a 2023 report on chatbots in consumer finance, the Consumer Financial Protection Bureau documented concerns about inaccurate information, privacy and security risks, and customers struggling to reach a human who could resolve their problems. That report examined financial customer service; it was not an assessment of the invitation I received. Its concerns remain relevant to the standards we should demand.
When money is involved, a mistake can have consequences well beyond an incorrect answer on a screen. Depending on the service and the authority granted, it could affect a payment, an investment decision, or a household’s ability to meet an obligation.
Any system permitted to take financial action should have safeguards that reflect those stakes. A conversational instruction such as “help me manage my money” should never be treated as blanket permission to decide where it goes.
My concern about regulation also needs precision. AI operates within a world of existing laws. In a 2023 joint statement, federal agencies emphasized that automated systems remain subject to the laws they enforce. Calling something artificial intelligence does not, by itself, exempt a company from consumer protection or civil rights obligations.
Still, the existence of legal protections does not answer every question about a particular product. Who is responsible for an error? How quickly can a consumer reach someone with the authority to fix it? What remedy is available? Are the safeguards independently examined, or simply described by the company selling the service?
Those questions need answers before consumers are asked to place substantial trust in a new financial intermediary.
If companies want that trust, they should make the terms understandable. They should explain whether access is limited to reading information or includes taking action. Any expansion of authority should require fresh, explicit consent. Financial actions should have clear approval requirements, meaningful limits, and records the customer can inspect.
Revoking access should be straightforward. The company should explain separately what disconnection stops and what it does with information already obtained. And when something goes wrong, a consumer should be able to reach a human being who can take responsibility.
These are the standards I want to see, not claims that every service already meets them.
We also need to respect the person who declines. An invitation to connect an account should remain an invitation. Choosing to keep financial information private is a legitimate decision, even when someone else finds the same service useful.
For my part, I can still ask AI to explain financial concepts, help structure a hypothetical budget, or work through figures I choose to provide without linking an account. Even then, I can limit sensitive details and verify consequential conclusions.
That is how I want to approach an augmented life: thoughtfully, with curiosity, and with boundaries I understand.
The hopeful future of AI is one in which people gain knowledge, confidence, and agency. We should be able to adopt useful tools at a pace that allows us to judge what we are giving them in return.
So when ChatGPT asked for access to my financial accounts, I said no.
That answer may change someday, if the access is clear, the protections are convincing, and the accountability is real.
For now, my money, my privacy, and my decision remain firmly my own.
Technology is changing the world. Read about it → The Augmented Life