top of page

SMA Crown Confidential Positioning Note: Control and Responsibility

Nov 8
2 min read

Updated: 2 days ago


Occasion: Wells Fargo published an updated Online Access Agreement, effective 24 November 2026 (confirmed: agreement published). The text that could be read treats transactions performed by a customer’s delegates, including their AI agents, as authorised by the customer, and makes the customer responsible for supervising them.


Position occasioned: on the client’s side of the table, lit from the client’s world: the bank keeps the control, the client carries the responsibility. The bespoke AI Confidante does not resolve that arrangement; it stays on the client’s side within limits it did not set. The institution is treated as a counterparty, not an opponent.


This week Wells Fargo published an updated online access agreement, to take effect on 24 November. In the text we could read, transactions performed by a customer’s delegates, their AI agents included, are treated as authorised by the customer, and the customer is responsible for supervising them. The bank also reserves the right to restrict or block certain outside tools. Some of the sharper wording has so far reached us only through secondary reporting, and we leave it aside.


We do not read this as hostile. A bank is entitled to say where its responsibility ends, and it is better that it says so in a contract than leaves it to be discovered. The institution is a counterparty, not an opponent, and a counterparty that states its terms plainly is easier to deal with than one that does not.


What the agreement does show is how the weight falls. The control stays with the bank: it may act on an agent’s instructions, and it may block the outside tools it chooses, and the text we could read does not say in advance which. The responsibility goes to the client: for every transaction their agents perform, and for supervising them. Nothing in that is wrong; the bank is the bank. But it leaves the client answerable for tools they cannot be sure they will be allowed to use, and likely to learn where the line falls only by meeting it.


A Bespoke AI Confidante does not resolve this arrangement, and we would not claim it does. It cannot lift a block, rewrite a term, or decide on the client’s behalf what to delegate. Nor does it need the bank’s permission to do what it does, because it never asks for access to the account at all. What it offers is narrower and steadier: it stays on the client’s side of the table, within limits it did not set, helping the client understand what they are answerable for, what they would rather keep in their own hands, and how to meet the bank’s terms with that understanding. Not a remedy. A presence on their side, whatever the bank decides to admit.


Founder & CEO of SMA Crown Confidential


Bespoke AI Intelligence for Private Decision-Makers




Comments


bottom of page