cd /news/ai-agents/citibank-updates-terms-holds-custome… · home › topics › ai-agents › article
[ARTICLE · art-171975] src=cryptobriefing.com ↗ pub= topic=ai-agents verified=true sentiment=· neutral

Citibank updates terms, holds customers liable for AI agent risks

Citibank changed its online user agreement this week to treat any access or transaction performed by a third-party AI agent as authorized by the customer, making customers liable for resulting losses even if they are hacked, with only limited legal exceptions. Citibank, the third largest US commercial bank, joins JPMorgan Chase and Bank of America, which revised their terms in late October 2026, and Wells Fargo, whose updated Online Access Agreement takes effect November 24, 2026, in restricting AI agent access to customer accounts. JPMorgan's revised Digital Services Agreement bars third parties including AI agents from accessing customer login information but still permits read-only access through regulated open-banking platforms.

by read3 min views1 publishedNov 18, 2026
Citibank updates terms, holds customers liable for AI agent risks
Image: Cryptobriefing (auto-discovered)

Photo: Tima Miroshnichenko / Pexels

The third largest US commercial bank joins JPMorgan and Bank of America in putting AI agent mishaps on the customer's tab

Citibank has rewritten its rules for the AI age, and the short version is this: if your AI assistant messes up your bank account, that is on you.

The third largest commercial bank in the United States changed its terms of service this week. Customers who hand their banking access to third-party AI tools now carry the risk if something goes wrong, even if they get hacked.

That puts Citi in step with JPMorgan and Bank of America, the two largest banks in the country, which rolled out similar policies last month.

What Citi actually changed #

The update came as a new section in Citi’s online user agreement. It deals specifically with third-party AI agents.

The core instruction is simple: customers should not share their login credentials. The more consequential part is what happens if they do.

Under the revised terms, any access or transactions carried out by an AI agent will be treated as authorized by the customer. The customer is held accountable for resulting problems, with only limited legal exceptions.

AI agents have become capable enough that this question was going to come up eventually. These tools range from models built by frontier AI labs to apps designed purely for personal finance.

Given a customer’s bank login, they can review spending patterns and weigh investment choices. They can also offer advice and, increasingly, act on it by moving money and paying bills.

AI, tech, and the markets they move—in one daily briefing.

Daily. Free. Join 34,000+ readers across crypto, finance, and policy.

The big banks move as a pack #

Citi is not freelancing here. Between mid-October and late October 2026, several of the largest US banks spelled out where liability sits when AI tools touch customer accounts.

JPMorgan Chase revised its Digital Services Agreement in late October 2026. The new language bars third parties, AI agents included, from accessing customer login information.

There is a notable carve-out, though. JPMorgan still allows read-only access through regulated open-banking platforms.

Bank of America took a similar line in late October 2026. It told customers that using AI tools with their credentials means they take responsibility for whatever happens in the account.

Wells Fargo followed with an updated Online Access Agreement, effective November 24, 2026. Its language classifies transactions initiated by AI agents as authorized by the customer.

Add Citi, and four of the country’s biggest banks are now restricting AI agent access to customer accounts. Executives at these institutions have also stressed the risk of unauthorized transactions by third-party agents and the importance of keeping a human in the loop.

Why the liability shift matters #

The phrase “considered authorized by the customer” carries a lot of weight. In banking, whether a transaction counts as authorized often decides who absorbs the loss when something goes sideways.

By defining agent activity as customer-authorized up front, banks are drawing a bright line before disputes ever arise. If an AI tool makes a bad payment, or gets compromised and drains an account, the bank’s terms already say who owns that outcome.

The pattern across the banks also hints at a preferred path forward. JPMorgan’s exception for read-only access through regulated open-banking platforms suggests the industry is not banning outside tools outright. It is steering them toward channels the banks can see and control.

Read-only access lets a tool analyze your finances without ever holding the keys to move money. Credential sharing gives the tool the same power you have, and that is precisely what the banks are trying to discourage.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.

── more in #ai-agents 4 stories · sorted by recency
── more on @citibank 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/citibank-updates-ter…] indexed:0 read:3min 2026-11-18 · —