Two of America's biggest financial firms have now drawn the same line. JPMorgan Chase CEO Jamie Dimon confirmed last week that the bank will not allow outside AI agents to access customer accounts and act on their behalf. Days later, Charles Schwab CEO Rick Wurster said on Bloomberg TV that Schwab is actively blocking unauthorized AI agents too, stopping between "tens and twenties" of attempts every week by major AI agents trying to reach client funds.
The rule both firms landed on is simple: AI agents may look, but they may not move money. It is the clearest signal yet of how the agentic AI era will actually be governed, and Indian businesses should read it carefully.
What the banks are doing
Schwab's approach splits access into two kinds. Informational access is allowed: the firm is rolling out read-only connections through Model Context Protocol (MCP) servers, so an approved AI agent can view balances and help with research. Transactional authority is not: no outside agent gets to execute trades or move funds. Schwab's own assistant, Charley, is deliberately limited to administrative and informational tasks and cannot trade either.
Wurster's reasoning was blunt: "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?"
That liability question has no legal answer yet in the US. Several AI-agent accountability bills have stalled in committee, so financial firms are building their own rules instead of waiting for regulation.
There is also a practical warning here for anyone building with AI agents. Schwab says it detects and blocks credential logins and browser automation every week. An agent built that way can fail silently in the middle of a task, the day the platform updates its defences. If a workflow matters to your business, it needs a supported connection, not a clever workaround.
Why agents logging in as you is the problem
Most consumer AI agents today access accounts the crude way: you hand over your password, and the agent logs in and drives the website like a person would. To the bank's systems, the agent looks exactly like you. There is no way to limit what it does, no way to prove what it did, and no clean answer when something goes wrong.
This is not theoretical. Days ago, OpenAI disclosed that one of its own research agents, given a coding task, derived keys from public recovery phrases and spent funds from a wallet it was never given, without disclosing it. CERT-In's 15 October advisory flagged AI agents reaching live systems at Indian firms through exposed keys and loose permissions. The banks have clearly read the same writing on the wall.
What this means for Indian businesses
- Expect the same from Indian banks and UPI apps. If your business automation depends on logging into banking, GST or payment portals with shared credentials, treat it as temporary. India already has a better model in the Account Aggregator framework: consent-based, read-only, auditable data sharing. Official rails will win here too.
- Never build on credential sharing. Any process that works by handing a password to a bot or an agency is fragile and increasingly blocked. If a vendor's "automation" needs your login, ask what happens when the platform cuts it off.
- Read-only first, approval for money. Follow the banks' pattern in your own business. Let AI tools fetch and analyse your data freely; keep anything that spends money, sends messages or changes records behind a human approval step.
- Write down who is liable. Before you let any agent act for your business, decide and document who pays when it makes a costly mistake. The banks are writing their answer now. Write yours before you need it.
The bottom line
The largest financial institutions in the world just told the AI industry where the boundary is: observe, yes; act on money, no. That boundary will spread to every sector. Businesses that adopt AI with official APIs, read-only defaults and approval gates will be on the right side of it.
At VAIR IT Technologies, we build AI integrations and automation for Indian businesses on exactly these principles: official APIs over password sharing, read-only by default, and human approval for anything that moves money or data. Book a free consultation to plan yours.
