Artificial intelligence tools have made financial advice more accessible, but they have also created a significant privacy risk.
Millions of Americans are turning to AI chatbots for help with their finances, asking about budgets, debt payoff plans, retirement strategies and investment options.
In many cases, AI can be useful. But here’s the problem: To get specific answers, people are sharing too much; the kind of data that, in the wrong hands, could empty a bank account or lead to identity theft.
A Cisco survey found that 29 percent of global AI users entered personal or confidential information — including financial and health details — into chatbots, even though 84 percent were concerned their data would be made public.
A Stanford study examining the privacy policies of six major AI companies — Amazon, Anthropic, Google, Meta, Microsoft and OpenAI — found that all six use chat data by default to train their models, and some keep this information indefinitely. (Amazon founder Jeff Bezos owns The Washington Post. The Post has a content partnership with OpenAI.)
As noted by that research, “chatbots are poised to fundamentally change how we disclose personal data not just for our information-seeking needs but also to find answers to highly personal and intimate questions.”
Here’s a cryptic warning from the researchers: Either by design or negligence, your data could be exploited.
A year ago, only 1 in 10 Americans said they were using AI to manage their finances, according to TD Bank.
But in its latest poll, TD Bank said that figure has jumped to 55 percent, and among younger adults, it’s even higher — 77 percent of Gen Z and 72 percent of millennials are using AI for financial guidance.
AI is best used for general financial information. Use it to confirm the contribution limits for your 401(k) or Roth IRA. You might ask about the pros and cons of various debt-reduction plans. Or ask it to create a healthy meal plan for your family of four with a grocery budget of about $200 per week.
Here are five categories of financial details AI chatbots have no business knowing, along with some tips.
Personally identifiable information
Never share your name, address, Social Security number or driver’s license number with AI. You may need a username or email to log in, but keep those details out of the chat log.
You may be thinking, “Who would input such information?”
People give it up all the time in other contexts, without considering the consequences, often believing they have no choice. Once, as I was picking up some furniture, the store clerk wanted to take my driver’s license and scan a copy into the retailer’s computer system. He insisted that the information was needed to maintain a record of the items customers picked up. “No, sir,” I said defiantly. However, I was okay with him verifying my identity by eyeballing my license, but I had no idea how secure their system was.
My dentist wanted to use my Social Security number as my identification for my file. “No, ma’am,” I told the receptionist. I have always advised pushing back on requests for your Social Security number when there is no clear reason for the company, or medical professional, to have it.
You should always be fiercely protective of your personal data, and this definitely applies to using AI.
Employment information
In the Cisco survey, 37 percent of respondents said they have entered work information into an AI tool.
Employment details are often the missing piece of the puzzle for phishing schemes.
If a scammer knows where you work, that’s a convincing detail that may allow the person to apply for a credit card or loan in your name. Or they could send you a fraudulent email asking you to click on a link to update your W-2.
These attacks often work because of the specificity.
The AI tool doesn’t need to know you work at a certain company to give you advice on negotiating a raise. Ask general questions that cannot be traced directly to you.
Specific debt data
When seeking financial advice from a chatbot, don’t share amounts owed to specific creditors.
You should not ask, “How should I prioritize paying off the following debt if I’m getting a tax refund of $3,450?” and then type, “I owe $14,200 to Citibank and $8,700 to Chase.”
Major data breaches have already exposed a great deal of our information. As a result, fraudsters can craft convincing schemes that are incredibly difficult to distinguish from legitimate communication.
If your chat is exposed in a breach, and a scamming operation can connect your conversation back to you, they may have enough to persuade you they’re calling from your lender.
Use round numbers if you’re asking questions about managing your money.
Specific transactional data
Maybe you’re asking AI to help prepare a budget for you. You think nothing of inputting what you spent last month: $45.12 at your local pharmacy or $65.23 for happy hour and dinner.
Or you might want specific advice and tips, so you upload your actual budget. What if that information is compromised?
In many scam calls, a fraudster pretending to be from your bank could disarm you by sharing purchase amounts. In the moment, you might think it’s a legitimate call because they have that information.
When you give an AI exact numbers, you move from an anonymous situation to an identifiable individual. If a data breach occurs or if the AI company’s data is sold, a bad actor could cross-reference your chat with leaked credit card metadata to figure out exactly who you are.
Financial documents
I get it. This money stuff is confusing and complex. You need help, so why not ask AI for assistance. You figure it saves time and money.
I can see the temptation to upload your tax return to get help with finding deductions. Perhaps, to protect yourself, you think blacking out your name and Social Security number makes it safer. But there is a lot of information on a tax return that could still lead back to your identity.
Or, maybe you don’t want to pay a financial adviser so you upload your investment statement and ask an AI tool to analyze your portfolio.
But just like you wouldn’t hand this information to a stranger, don’t give it to AI. Ever.
Keep your AI chat generic
An AI chatbot does not need your account number to tell you how to pay down debt, nor does it need your Social Security estimated earnings statement to recommend when to start collecting your retirement benefit.
The less you share with AI, the safer your money will be.
Look for the privacy setting on the platform you are using and opt out of having your chats be used for model training.
Keep the identifying details out of your chats. With AI, as with most online contact or conversations, trust nothing.
The post Don’t tell your AI chatbot these 5 things to keep your money safe appeared first on Washington Post.
