Would you accept financial advice from a sociopath? If you’re among the majority of Americans using AI to manage their money, chances are you’ve already done so.
MIT finance professor Andrew Lo co-authored a 2023 paper that equated large language models (LLMs) like ChatGPT to “a human sociopath” due to their inability to express empathy (1).
However, that doesn’t seem to bother the 55% of Americans who said in a March 2026 TD Bank survey that they “use AI to help manage their finances (2).” This represents an increase compared to 10% the previous year.
Must read
Of course, the use of AI for financial assistance is not new. Americans turn to him for everything from financial literacy questions to savings plans, stock market advice, and retirement advice (3). Lo himself has talked about how to use AI as a financial advisor. However, the growing concern around artificial intelligence and finance has more to do with the information people provide to LLMs to get the advice they seek.
A 2024 Cisco survey confirmed that up to 29% of AI users enter account numbers and other financial information, even though the vast majority acknowledge that their data could be shared (4).
But data sharing is not the only concern. Thieves have discovered devious ways to obtain your personal information from LLMs, data they can then use to steal your identity and money.
How criminals steal your data from AI tools
Last year, researchers at Stanford University studied six major US LLMs: Nova, Claude, Gemini, Meta AI, Copilot and ChatGPT (5).
Lead author Jennifer King said they found that any “sensitive information” shared with such LLMs “can be collected and used for training” — a big problem, she added, because “almost no research has been done to examine the privacy practices of these emerging tools.”
NordPass, a division of Nord Security, warned that security breaches in LLMs may allow “malicious actors” “to view your entire chat history, including any sensitive data you have shared with the AI tool (6).”
Others, meanwhile, point out that the personal information you upload is used to train AI models, which then “could reproduce text word for word, exposing personal information (7).”
Cybercriminals can also steal your data through a process called indirect notice injection, where they hide notices for LLM inside anything from web pages to images, emails and documents (8).
Imagine you are using an LLM and you unknowingly upload a web link or document that has a hidden injection inside it, one that tells the LLM to ignore your request and instead reveal your password and other sensitive information to the cybercriminal who set the trap.
The LLM will complete the request to “expose sensitive data, trigger unwanted actions, or bypass security filters,” according to Norton (9). That data breach could include financial information that you entrusted to the AI tool.
Once cybercriminals have your financial information, they can do everything from steal your identity, borrow money, and open accounts in your name to simply selling your data to other malicious actors on the dark web (10).
Read more: This billion-dollar private real estate fund is now accessible to non-millionaires. Start investing with just $10
What you should never reveal to AI
Most of the information that experts advise never sharing with AI seems like common sense, but it’s worth repeating.
Always avoid sharing personal financial information, including bank account numbers, credit cards and Social Security numbers, passwords, your banking institutions, and even specific dollar amounts, such as the amount of debt you want to pay off.
“If your chat gets exposed in a breach,” the Washington Post said, “and a scam operation can connect their conversation to you, they may have enough to convince you they’re calling from your lender (11).”
Others recommend keeping your name and address, phone number, birthday, and other general personal information out of LLMs, along with medical records and work-related data (12).
Norton even warned that entering creative work, such as a script you wrote and need help editing, could result in that content being replicated in other searches, stripping you of ownership and “leaving you vulnerable to legal and financial risks.”
To be safe, never share your real name or legal or identifying information with an AI tool, and always check the URL to make sure you’re on a legitimate LLM site rather than a malicious imposter (13).
Additionally, if the LLM has a privacy setting that allows you to prevent sharing of your data, enable it (14).
If you need to ask questions about personal matters, use made-up but realistic-sounding data and financial information so you can get approximate answers without revealing anything specific about yourself (15). Then delete your chat history, just to be safe.
Lastly, don’t forget to verify any LLM information or advice with a legitimate source. Remember, after all, they are sociopaths.
you may also like
Join over 250,000 readers and get the best Moneywise exclusive stories and interviews first – clear insights curated and delivered weekly. Subscribe now.
Article sources
We rely only on verified sources and credible third-party reports. For more details, see our ethics and guidelines.
MIT GenAI (1); TD Bank (2); Credit Karma (3); Cisco (4); Stanford University (5); NordPass (6),(13); Milvus (7); EC Council (8); Norton (9),(14); Aura (10); The Washington Post (11), (12); Fixed password (15)
This article originally appeared on Moneywise.com with the title: More Americans are asking AI for financial advice and revealing too much personal information. What you should not share
This article provides information only and should not be construed as advice. It is provided without warranty of any kind.
