Skip to content
Rush Commerce
AI & Automation3 min read

Citi, PNC: you're liable for what your AI agent does at the bank

Citi, PNC and Citizens changed terms: if you give an AI agent bank access, you own its actions, even if the tool is hacked. Keep bank credentials out of AI tools.

If you give an AI agent your bank login, the bank now says every result is yours. This week Citi changed its Online User Agreement: customers who grant account access to third-party AI agents are responsible for what those agents do. PNC changed its terms on November 18. Citizens emailed customers a similar warning. Citi's update goes one step further. If the AI tool leaks your credentials or is hacked, you are responsible for what the attackers do too, TechCrunch reports.

What actually happened

Per TechCrunch:

  • Citi updated its Online User Agreement. A spokesperson said customers "who choose to grant account access to third-party AI agents are responsible for activities conducted through those agents." The terms also cover credential leaks and hacks of the AI tool.
  • JPMorgan and Bank of America announced similar policies last month.
  • PNC changed its terms on November 18. Citizens sent customers an email this week with the same message.
  • The exception is the bank's own agent. Citi's terms say customers can use AI agents that Citi provides without the added liability.
  • Not every AI vendor wants your login. Anthropic's new personal finance tool is billed as read-only. It "can't move, manage, or invest your money."

TechCrunch's report covers consumer accounts and the online user agreements. It does not say whether business accounts have the same terms. Read your own bank's business agreement before you assume either way.

In October, we wrote that banks were blocking third-party AI agents and that read-only access was the likely middle path. This is the next step. The banks no longer only block agents. They put the risk on you in writing.

Why it matters for your business

Many small teams already paste a bank login into a bookkeeping bot, a cash-flow app or a general AI agent. Under these terms, a wrong transfer is your loss. A breach at that AI vendor is also your loss. The bank has no duty to reverse it.

What we would do this week:

  1. Find every tool that holds a bank login. Bookkeeping apps, expense tools, AI assistants, browser agents. Write them down.
  2. Move to read-only feeds. Use the bank's own data connection or an aggregator with read-only scopes. A read-only feed lets the agent analyze. It can't move money.
  3. Keep payments behind a human. Let the agent draft the bill run. A person approves it in the bank's own portal.
  4. Rotate the passwords that any AI tool has seen.

Key takeaways

  • Citi, PNC and Citizens now tell customers they own everything a third-party AI agent does in their account
  • Citi's terms also make you responsible if the AI tool is hacked or leaks your credentials
  • JPMorgan and Bank of America set similar policies in October; the bank's own AI agents are the exception
  • Coverage is about consumer terms; check your business account agreement yourself
  • Give AI tools read-only bank data and keep money movement in the bank's portal

Want AI on your books without handing it the keys? We build finance automations on read-only data, with a human approval step before any money moves. See what we build or tell us which tools touch your bank today.

Sources: TechCrunch.

  • #ai-agents
  • #banking
  • #agent-security
  • #liability
  • #finance-automation
TR

Tommy Rush — Founder, Rush Commerce

Operator turned builder. 15+ years running operations — now shipping the systems businesses run on. More

Get The Rush Report weekly — one email, zero fluff.