News

Citi becomes latest US bank to restrict AI agents from accessing accounts

Citi has joined JPMorgan in restricting AI agents. It does not permit 'third-party AI agents or assistants to log into a customer's account or transact directly', though read-only access through certain open-banking providers is still allowed. Bank of America is also restricting agents: CEO Brian Moynihan told CNBC he wants 'a human in the loop', and JPMorgan's Jamie Dimon cited 'rogue agents'. The trigger was Meta's Muse personal AI assistant, launched in September. Banks fear agents could sweep idle cash out of checking accounts paying an average 0.1% into accounts paying 3.3–5% (Apollo's Torsten Slok). JPMorgan, BofA and Wells Fargo hold $1.6tn of non-interest-bearing deposits (16% of liabilities), and paying 3% on that would cost $47bn a year. Executives have contacted the Federal Reserve, and regulators have privately welcomed the blocks. Who is liable for agent-initiated payments is still unresolved. Caveats: the Fed declined to comment, and Citi frames the policy as safeguarding accounts. Simon-Kucher's Christoph Stegmeier says banks' answer should be to 'make your deposit products stickier'.

Why it matters

This is the first concrete incumbent pushback against agentic finance. The biggest US banks are shutting autonomous agents out of transactions while keeping read-only open banking. That sets the terms for agent-payments, deposit-sweep and open-banking startups, and it is a template European banks could follow. The deposit-cost maths ($1.6tn, $47bn a year) shows what is at stake. It is separate from the Daily Deal story about an OpenAI agent spending from a stranger's crypto wallet.

Read the full article: Financial Times

More top stories