Compare your stack: agent access to customer accounts
An eight-control self-check for bank and credit-union security leaders, built on the Nov 9, 2026 edition. Five big UK lenders have moved to restrict AI agents in customer accounts (The Telegraph, Nov 8), and the OCC's Nov 4 examiner framework rates AI by consequence, transfer of agency and human oversight. Score yourself, then see what an examiner-ready answer looks like.
For: CISO, head of digital channels, head of fraudWhere peers drew the line
| Who | Reported move | Date and source |
|---|---|---|
| HSBC, Barclays, Lloyds, NatWest, Nationwide | Moved to block AI agents from customer accounts; some ban AI-initiated transactions without the customer's direct involvement | Nov 8, 2026 · The Telegraph via Yahoo Finance (CITED) |
| JPMorganChase, Citi, Bank of America, Wells Fargo, Morgan Stanley | Block external AI agents from customer accounts | Oct 27, 2026 · American Banker, as tallied by FourWeekMBA (CITED) |
| Charles Schwab | Blocks 10 to 20 clients a week from using AI agents (CEO) | Nov 8, 2026 · The Telegraph (CITED) |
Score your controls
Tick what is true today. The score stays in your browser.
0 of 8
Why the examiner will ask
OCC Bulletin 2026-54 (Nov 4, 2026) tells examiners to classify use cases "by potential consequences, transfer of agency, and degree of human oversight," rather than technical sophistication. It is an examiner resource and "does not establish supervisory expectations for banks" (VERIFIED, occ.gov). Summaries report it flags transaction-initiating agents, one-click approvals, and banks' rights to "monitor, test, restrict, or terminate" third-party systems (CITED, Consumer Finance Monitor, Nov 6, 2026). Comments on the interagency third-party risk guidance close Nov 16, 2026 (VERIFIED, Federal Reserve).
Our take (opinion): block first if you must, but publish a narrow path: read-only through consented APIs, initiation only with in-app customer approval, and a revocation switch you own.