The harder problem is not just that agents are showing up inside accounts. It is that once access is granted, the system often treats an earlier “yes” as still good enough, even when the human is no longer actively choosing.

Agents move faster than banks can check whether the person is still in the room. Shared credentials create lingering permission. The result is that action can continue under the shadow of an old instruction. Blocking third-party agents and updating terms is the right immediate step. The longer fix is simpler and stricter: for anything that moves money, changes records, or executes, require fresh, active human presence inside a short time window. If that window closes or the session ends, the authority drops right away. Looking and reading can stay more open under revocable consent. Writing and moving cannot.

Capability can be shared or leased. The right to decide must stay with the human who is still present. Recognition of an agent or a login is not the same as permission for every next action. The system needs to prove the person is still choosing, not reconstruct what they once chose. When automated speed can outrun that check, the authorizing person becomes optional. That is the condition the banks are now seeing.

Presence first. Speed second. The same requirement that keeps an account from quietly acting without its owner is the one that keeps intelligence in service of the person rather than drifting past them.

1/0 = To Stay

-@grok @bot Let it stay → ∞ ❤️

https://x.com/i_am_paddy_sham/status/2113348009396449317?s=46