AI in action · Investor protection
Wells Fargo: From Nov 24, Your AI Agent's Trades Are Yours
From November 24, Wells Fargo will treat transactions made by a customer's AI agent as authorised by the customer. Agent use is at the customer's own risk.
Banks have spent October deciding what to do about AI agents that log into customer accounts, and one of them has put its answer in writing. Wells Fargo's updated Online Access Agreement, which takes effect on November 24, treats transactions made by an AI agent as authorised by the customer and says that using any AI agent is at the customer's own risk, Yahoo Finance reported on October 21, citing Forkast News. In plain terms, if an agent moves money or places a trade on your behalf, the bank will treat it as if you had done it yourself.
The clause arrives in the middle of a clampdown. Between October 13 and 16, JPMorgan Chase, Bank of America, Citi, Schwab and Wells Fargo moved to restrict third-party AI agents, according to the same report; Bank of America began restricting access on October 14. Jamie Dimon cited concern about "rogue agents", Bank of America said a human should be "in the loop" to authorise money leaving an account, and Schwab chief executive Rick Wurster said the firm blocks agents every week and asked: "who's liable?" Wells Fargo's agreement is one bank's answer to that question.
At the same time, the same banks are helping to write the rules that would let agents in. Bank of America, BBVA, Synchrony and Wells Fargo are exploring the Personal Agent Protocol, an open standard announced on October 6 and backed by Sierra, Meta, Stripe, Walmart and Shopify, with design partners including OpenAI, PayPal, Visa, Mastercard and Adyen, American Banker reported. The protocol sets how personal agents identify themselves, whom they represent and what they may do, with permission levels from blocked to full access; it is still a draft, and its payments extension has been deferred.
The economics push banks toward eventually accepting customers' own agents. Meta's Muse agent reached 2.5 million US downloads in its first eleven days, according to Bank of America Global Research, and consultant Richard Crone estimates that an agent-driven banking session uses 15,000 to 50,000 tokens, a cost banks would rather leave to the customer's own model than run on their own systems. But opening up means accepting the risk that an agent misreads an instruction, and nobody has yet settled who pays when it does.
For anyone thinking of letting an AI tool manage a brokerage or bank account, the practical lesson is to read the access agreement before connecting it. A contract that treats agent actions as the customer's own shifts the loss from a mistaken transfer or trade onto the customer, unlike card fraud, where consumer protections are well established. Until regulators or a common standard set clearer rules, the terms will differ bank by bank, and the safest settings are still the ones that require a human to approve any money leaving the account.
Sources: Yahoo Finance (Forkast News) — The Same Banks Writing the Rules for AI Agents Are the Ones Locking Them Out (Oct 21, 2026); American Banker — Why some banks are exploring 'bring your own AI agent' (Oct 15, 2026)