Why big banks are blocking AI agents, and what kind of AI agent a business will let in

In the same week of October 2026, two big US financial firms said much the same thing: outside AI agents can’t just log in to customer accounts and act for them.

That matters even if you never touch a US bank. It shows where the line is being drawn for AI agents, and it gives a small business a simple test for any AI tool it lets near its customers, money or chats.

What JPMorgan said

On 13 October 2026, JPMorgan Chase CEO Jamie Dimon told analysts on the bank’s third-quarter earnings call: “We’re not going to allow just any … outside agent to come in inside the company, because it’s just too risky for clients and for JPMorgan,” as reported by CNBC.

A JPMorgan spokeswoman told CNBC: “we do not permit third parties to access accounts by obtaining customers’ banking credentials and logging in on their behalf.” She added that the bank supports “secure, customer-authorized data sharing through open banking providers.” CNBC reports this means AI agents can see some Chase account data in a read-only way, such as balances, if Plaid or another connector has the user’s credentials, but they can’t send money or pay a bill.

What Schwab said

On 15 October 2026, on Schwab’s Fall Business Update call for investors, CEO Rick Wurster said, as reported by American Banker: “there’s one way we think is not safe, and that’s just an agent coming to our site and trying to log in on behalf of a client. We are not going to allow that.” American Banker reports Schwab is stopping agents that use clients’ login details or browser automation.

The next day he told Bloomberg Television, as reported by Moneycontrol, citing Bloomberg: “We don’t want someone just giving their password to Muse, and letting Muse come in and take action on that client’s behalf. We can’t control it, the client can’t control it, something could go wrong and who’s liable?” Muse is an AI agent from Meta.

Schwab isn’t banning AI. According to the same report, Wurster said Schwab will prepare model context protocol (MCP) servers to connect outside AI tools to Schwab’s systems. MCP is a set of rules for linking AI tools to other systems. “What that allows us to do is control the authentication process and also control what they can access,” he said. He said clients will be able to see their account balances and do research that way, but not move money.

The real line: logging in as you vs. a limited permission

Both firms drew the same line:

• Not allowed: an AI agent that holds your password, logs in as you, and can do anything you can do.
• Allowed: an AI tool that connects through an official, approved route, with limited permission that the business controls.

The MCP rules show why the second way is safer. MCP’s official authorization spec says its sign-in method is based on OAuth 2.1, a standard often behind the “Sign in with…” approvals you see on many apps. (The spec makes authorization optional, so not every MCP connection uses it.) When it is used, the tool gets an access token for one specific service instead of your password, and the spec says sign-in servers should issue short-lived tokens. Done this way, a business can limit what the tool can do and cut it off without changing your password.

Why this matters for a business on WhatsApp

The same idea applies to the AI that answers your customers on WhatsApp.

WhatsApp already draws this line. Its Help Center says linking your account to unofficial versions of WhatsApp “violates our Terms of Service,” and your account “might also be temporarily or permanently banned.” WhatsApp’s Terms also ban bulk messaging, auto-messaging and unauthorised automated access. For businesses that want to connect software, Meta offers the WhatsApp Business Platform, which it describes as its business messaging APIs.

So, like the banks, the question isn’t “AI or no AI.” It’s “which door does the AI come through?”

Five questions to ask any AI agent vendor

1. How does it connect? Through an official API or approved connector, or by logging in with your password or copying a browser session? Avoid the second.
2. What can it do, and what can’t it? Ask for the list. A good setup can read and reply but needs a person for money, refunds and bank-detail changes. It’s a similar line to the one JPMorgan and Schwab drew on moving money.
3. Can you switch it off on its own? You should be able to remove its access without changing your own passwords.
4. Is there a record? You should be able to see what it said and did, and when.
5. Who answers when it goes wrong? Wurster’s question, “who’s liable?”, is a fair one to ask any vendor before you sign.

Where ThinkChat fits

At ThinkChat, we think a good AI CS employee answers from information you approved, stays honest when it doesn’t know, and can hand over to a person. Ask us the five questions above too, the same way you’d ask any vendor.

Soft next step

If you want to hire a ThinkChat AI CS employee, Talk to Lyia on WhatsApp: +60 11-5515 9318.

Sources

Checked 21 October 2026.

• CNBC, “JPMorgan Chase blocks outside AI agents like Muse, Instinct from directly accessing accounts” (13 Oct 2026)
• American Banker, “Schwab blocks unauthorized AI agents, points clients to ‘Charley’” (15 Oct 2026)
• Moneycontrol, citing Bloomberg, “After JPMorgan, Charles Schwab blocks unauthorised AI agents from accessing customer accounts” (16 Oct 2026)
• Charles Schwab press release, “Schwab Delivers Record Net New Assets & Earnings” (15 Oct 2026)
• Model Context Protocol specification, “Authorization” (2025-06-18)
• WhatsApp Help Center, “About unofficial apps”
• WhatsApp Terms of Service
• WhatsApp Business Platform