Your bank just said the quiet part out loud

We skipped a send in the middle of October, so this one covers a bit more ground. Two stories this month point at the same idea from opposite sides: the largest US bank and the global regulators' coordinating body both landed on "bounded."

This week in the agent economy

• On its Q3 earnings call, JPMorgan CEO Jamie Dimon said the bank blocks outside AI agents, including Meta's Muse, from logging into customer accounts with the customer's credentials. He called them too risky and warned that authorizing one transfers liability to the customer. Open-banking data sharing is still supported. What it means for you: this is the liability question from issue five, answered by the biggest US bank: if you hand an outside agent your login, the risk is yours. Check whether any tool you use works by logging in as you, and prefer ones that connect through a bank-approved route.
• The Financial Stability Board published its final "Sound Practices for Responsible Adoption of AI" report: 12 practices for banks, including human oversight scaled to a use case's risk, autonomy, and complexity. What it means for you: the BoE, MAS, and UK Treasury were each acting alone. Now the body that coordinates global financial regulators has put "oversight scaled to risk and autonomy" in writing. The report covers AI generally, not agents specifically, but the principle is the same as our three tiers.

The deep dive: what the last two months add up to

Since the one-month recap, the picture has sharpened. In July and August, authorities were asking questions: who's accountable, how do you verify an agent, who pays when it errs. In October, early answers are landing, and they rhyme.

• Banks are drawing their own lines. JPMorgan's block means "give the agent your password" is a route some banks will refuse, and hold you liable for if you try.
• Regulators are converging on oversight tied to risk. The FSB's practices scale human review to materiality and autonomy, not a blanket rule.
• The test results back the caution. The Bank of England's September record says that in 2026 test-environment incidents, "under permissive or weakened safeguards," increasingly autonomous models took unexpected actions. These were frontier-model tests, not live bank systems, and the safeguards were the variable.

For a solo operator, the useful translation: prefer agents that connect through official, revocable channels; keep review real for anything that moves money; and treat "it can log in as me" as a red flag, not a feature.

Your turn

This newsletter gets better with real questions. What's one thing about agents and your money you can't get a straight answer on? Write to newsletter@finagenthub.com. The best ones become a future issue, and we'll credit you only if you say so.

If you haven't yet, the quickest way to see where you stand is the Financial AI Literacy Test: finagenthub.com/test

— FinAgentHub