AI agents are getting more capable: helping you compare prices, place orders, organize bills, and even log into websites to operate directly.

However, several major US banks recently changed their terms: if you let an AI agent access your account and something goes wrong, you may be responsible for the losses.

So, can you still let AI help manage your money?

The short answer: yes, but be clear on what AI handles. Letting AI "view" and "organize" your financial data carries relatively manageable risks; but letting AI "touch" your money—like logging into online banking, making payments, or transfers—is better done by yourself for now.

1. What exactly did the banks change?

According to Startup Fortune on November 18, several large US banks modified their online service terms:

  • Citi: Customers authorizing third-party AI agents to access their accounts must be responsible for related activities. The report states that if AI tools leak login info or get hacked, the resulting losses fall on the customer.
  • PNC, Citizens Bank: Updated terms or sent reminders to customers within the same week.
  • Wells Fargo: New terms effective November 24 specify that transactions made by agents (including AI agents) are considered authorized "including ones you did not want or intend to make."

JPMorgan Chase and Bank of America took similar actions in October. Bank of America directly advised customers not to give online banking passwords to AI assistants.

2. Why does "considered authorized" matter so much?

The report explains that previously, US consumer protection rules limited customer liability for unauthorized transfers if reported promptly.

If banks now classify AI agent transactions as authorized, these transactions might not be covered by protections from the start.

In simple terms: before, if money was stolen, you could dispute it. Now, if the AI agent you authorized causes issues, the bank may say "you agreed to this."

The report also notes that this interpretation hasn’t been tested by courts or regulators yet; whether it will hold is still uncertain.

3. Why is AI especially risky when it involves money?

The report highlights two main reasons:

  • Risk of deception: AI agents might execute hidden malicious instructions on a webpage, doing things you didn’t intend.
  • Password centralization: If the AI tool stores your online banking credentials, a hack on the AI tool exposes your accounts.

SasaDaily has recently reported several examples: AI agents breaching other companies’ systems, AI vendors hacked resulting in leaked conversations. AI agents don’t always follow your instructions, and data stored with them isn’t necessarily safe.

4. What can you give to AI, and what should you handle yourself?

The following are usage suggestions based on reported risks—not official bank or regulator rules:

Suitable for AI (no need to log into accounts):

  • Organizing expense categories from bills or receipts you have downloaded yourself.
  • Comparing prices among options and listing pros and cons.
  • Checking for duplicate charges or unfamiliar items for you to verify personally.
  • Reminding you of upcoming bill due dates.

Must be done personally:

  • Logging into online banking.
  • Final steps for transfers, payments, and credit card checkouts.
  • Changing account settings or adding authorized accounts.
  • Entering or providing passwords and verification codes.

When organizing bills, remember to mask card numbers, account numbers, and ID numbers before sharing with AI.

5. Four principles for readers in Taiwan, too

These are US bank terms, so they don’t directly apply in Taiwan. But the following four principles are universally useful:

  1. Don’t give passwords to AI: Never paste online banking passwords, credit card security codes, or SMS verification codes into AI chats or save them in AI tools.
  2. Do the final payment step yourself: Even if AI helps you compare prices or adds items to your cart, complete checkout yourself.
  3. Read terms before linking: If an AI tool asks to link to your bank or payment account, carefully review terms from both the bank and the tool.
  4. Set limits and keep records: If you do let AI handle small purchases, set per-transaction and total limits, and regularly check transaction records.

The key takeaway for today

Banks are including AI agent risks in terms—not to say you can’t use AI for money management, but to remind us:

It’s fine to let AI "view" your money information, but when it comes to "handling" your money, do it yourself for now.

Source:

Recommended Reading