Citi Restricts AI Agents from Accessing Customer Accounts

Summary

Citi is restricting AI agents from accessing customer accounts, according to the Financial Times, as banks become more cautious about software that can move money autonomously. Citi CEO Jane Fraser said at the 2026 Sibos conference that agents must undergo rigorous identity verification and authentication before executing transactions, whether they act for a customer or for the bank. Citi requires every internal AI agent to pass through ARC, an internal platform introduced in 2026 to monitor and govern agents. The bank has been gradually deploying governed AI systems internally since at least mid-2025, with compliance and oversight as central priorities. Fraser said she was confident Citi could avoid the risks of uncontrolled agents, and emphasized that the main concern is unverified agents rather than agents in general. JPMorgan Chase has taken a more restrictive public position: CEO Jamie Dimon said the bank would impose a total ban on external AI agents accessing JPMorgan customer accounts because of the danger of rogue agents. Citi has not announced an equivalent blanket ban on outside agents, instead emphasizing internal controls. The differing approaches could hinder personal-finance companies whose products depend on agents seeing and acting on users’ bank accounts. Citi’s governance model also requires staff, systems, and ongoing monitoring, making controlled agent access an operational cost as well as a security requirement.