
Key takeaways
- Morgan Stanley blocks third-party super intelligence (SI) agents from its platforms, while JPMorgan, Bank of America and Citi restrict them.
- Bank executives call unauthorized agents inside customer accounts horrifying, and Axios sources say most banks move toward read-only access.
- L402 pairs a Lightning payment with a macaroon, letting a machine pay and authenticate without a human's bank login.
Major banks in the United States are detecting unauthorized third-party super intelligence (SI) agents operating inside customer accounts, with some agents attempting to take actions that appear independent of user instructions, sources told Axios. Morgan Stanley has begun blocking third-party SI agents from accessing its platforms, while JPMorgan, Bank of America and Citi are also restricting agents, according to people familiar with the decisions.
Bank executives described the situation as horrifying, with one official saying the industry has grave concerns about agents gaining access to customer accounts and taking steps that could impact user finances.
Logins Handed to Bots
The agents generally access a human's account with the user's permission and login credentials, which goes against the guidelines of most banks requiring that humans log in to their own accounts. Some banks have also tracked agents contacting call centers or appearing to take actions independent of user instructions.
Many banks are broadly not able to parse whether agent account activity is user-directed, and one bank executive told Axios they struggle to tell whether an agent has been authorized by a user or has gone rogue. That difficulty is forcing banks to block the technology and rapidly rewrite their terms of service.
Read-Only Access and Denials
A Morgan Stanley spokesperson said the bank hasn't observed instances of "rogue AI agents accessing our platforms." A Chase spokesperson said customers should not give their usernames and passwords to anyone.
Most banks have authorized or are moving toward read-only access, which lets agents see customer financial information without taking action, the same kind of access banks already grant to third-party apps like Quicken. Executives told Axios they recognize customers will want to use agents, but for now they see the risks of being too permissive as too high.
Lightning Lets Machines Pay
Bitcoiners build machine payments instead. L402 is a protocol in which a server answers an unpaid request with a token and a Lightning invoice. After the client pays, it presents the macaroon, a bearer token, with the payment preimage, and the macaroon's cryptographic validity proves the payer obtained that preimage.
L402 pairs a Lightning payment with a macaroon, so a machine can pay and authenticate in the same exchange. L402 tokens are bearer instruments that can be restricted, for example to read-only access.
Why It Matters
Banks built their gatekeeping around a human typing a password, so when agents arrive with borrowed login credentials, Morgan Stanley blocks them and its rivals restrict them toward read-only access. That is the fiat permission model in miniature, where every action needs an intermediary's approval and anything the gatekeeper cannot verify gets locked out. Bitcoin embraces agents while banks block them, because L402 pairs a Lightning payment with a macaroon so a machine can pay and authenticate without human credentials or a bank's permission. We think the incentive story is plain: banks protect their liability and their gatekeeper role, while Bitcoin settlement lets software hold its own keys and pay its own way.





















