Ai Gives You Answers. But Who Takes The Blame?

It's fast, it's free, and it's genuinely useful. It also isn't on the hook.

If you're a physician, I'm sure you've had this patient.

They come in with their phone out, and they've already figured out what's wrong. They asked ChatGPT or WebMD, maybe even identified the medication. Sometimes they're right. Often they're close. But every so often, they're confidently and completely wrong in a way only someone with the right training would catch. My wife sees this often with her own patients.

The stakes can be life or death. In October the New York Times reported on a lawsuit alleging that ChatGPT walked a young woman with epilepsy through tapering off her seizure medication without her doctor's guidance, and she died as a result. Those claims haven't been proven in court. But even when the chatbot is right, finding a plausible answer isn't the same as someone taking responsibility for the patient’s health.

I'm starting to see the financial version of that patient. More and more people are asking AI to be their financial advisor.

I get why. I use AI every day myself, including to build software. So this isn't an anti-AI post. It's about what happens when things go wrong, and who's left holding the bag.

The largest bank in America just weighed in

On October 13, during JPMorgan Chase's earnings call, Jamie Dimon was asked about customers giving AI agents access to their accounts. His answer was blunt: "There's a liability transfer." His CFO, Jeremy Barnum, said customers who share their banking credentials with AI agents could bear responsibility if those AI agents take actions they didn't want. Chase now lets outside AI agents see a balance, but not send money or pay bills.

There's another side to that warning. Chase's standard savings account pays 0.01%. Plenty of online high-yield accounts pay over 4%. An AI agent that sweeps idle cash to the best rate threatens that spread, which is a big part of how banks make money. Both things can be true: the risk to you is real, and the bank has its own reasons for saying so. You don't need an AI agent to close that gap, though. Moving idle cash yourself takes less than five minutes.

Here's why the risk transfers to the customer. Federal rules known as Regulation E protect you from unauthorized transfers out of your bank account. But if you give someone access and authority to move money, transfers they make generally don't count as unauthorized, even if they go further than you intended. Give an AI tool your login, and in the bank's eyes you may have signed off on whatever it does next.

A hacker using stolen credentials is different, as you'd likely still be protected. But no regulator or court has spelled out how this applies to AI agents, and I wouldn't want my family's savings to be the test case.

When the advice is wrong

There’s a similar risk-transfer when it comes to advice. I'm held to a fiduciary standard, I'm registered, and if I get something wrong there's a regulator and my reputation at risk. ChatGPT, Claude, and Gemini aren't registered investment advisers and don't owe you the fiduciary standard. Their terms of service limit the company's liability and say, in effect, that what they tell you isn't financial advice. What you have is a well-crafted response.

These models generate the most probable answer, not a verified one. That's why a person who knows the field still needs to check the work, especially in medicine and money, where the stakes are serious. A well-written answer can still be wrong for you. Ask an AI how to do a backdoor Roth IRA and you'll get clean, accurate steps, perhaps even a nod to the pro-rata rule. What it can't know, unless you tell it, is that you rolled an old 401(k) into an IRA six years ago and forgot about it. Follow the steps perfectly and you can still end up with a tax bill you didn't see coming.

The things that matter most are usually the ones you didn't think to mention. That's how my parents overpaid their taxes, which I wrote about in "The $40,000 Nobody Was Looking For."

Context is the hard part

To be fair, an advisor also only knows what they learn about you. The difference is supposed to be what they do to learn it.

AI tools can now remember past conversations and read files you upload. What you can't assume is that the picture is complete, or that the tool will notice what's missing. It doesn't know your marriage is under strain, or that a parent is starting to repeat themselves. It won't pull the statements you didn't think mattered.

And much of what it learned from has already been edited. When I was in consulting, publishing was part of the job at my level. Most of it went out under a partner's name, scrubbed of anything too sensitive or useful to a competitor. The watered-down version that survived is what AI learns from, not the work done behind closed doors.

That's why we don't rely on what a client remembers, or treat what an AI model spits out as fact. We pull the statements, reconcile them against the tax return, look back a few years for what got missed, and come back to the plan every year as life changes. It's tedious, and it's exactly where the expensive mistakes hide.

How I'd actually use it

I'm not telling you to avoid AI. I'd be a hypocrite if I did.

Use it to learn. Ask it to explain a Roth conversion or a term in your benefits packet. It's an excellent tutor. Use it to prepare questions for your advisor, your CPA, or your attorney.

Don't give it your bank or brokerage login. A read-only connection through an approved service is lower-risk. Revoke access you're not using, and check whether the tool trains itself on your conversations. Many do, free or paid, unless you turn it off.

And don't act on its advice without someone accountable checking it, especially when the decision is big or hard to undo.

Someone on the hook

Knowledge was never the scarce thing. AI makes it nearly free. What's scarce is accountability.

That's the part of the work Sanjay and I care about most. We don't receive commissions, revenue-sharing, or incentives from fund or insurance companies; and our fee is the same no matter what we recommend. When something goes wrong, you know exactly whom to call. No prompts or AI call centers, a direct line to our cell phones. 

If you've been leaning on AI for your finances and aren't sure what it might be missing, reach out to us. The first conversation is always free, and it's exactly that: a conversation about your goals, what you've already figured out, and whether we can help.