America’s largest banks are facing a new problem they do not yet know how to handle. They have detected AI agents in their clients’ accounts that independently perform various tasks on behalf of humans. The banks have never vetted or approved these tools. According to findings by Axios, some of them attempted to take actions that users apparently never asked them to perform. Bankers are calling it a nightmare and say the technology is entering banking systems faster than they can respond to it.
Clients Let Them Into Their Accounts
At first glance, it might seem like a hacker attack. In most cases, however, the situation is different, because customers themselves are letting agents into their accounts. They give their AI assistant the username and password for their online banking so that it can, for example, monitor spending, pay bills, or transfer money on their behalf.
The client agrees to such an arrangement, but the bank does not. The rules of most financial institutions assume that only the account holder personally logs into the account. Banks’ greatest concern now is not artificial intelligence searching for vulnerabilities in their systems. They are far more troubled by agents that customers themselves have given access to their money.
Agents Call Customer Support
Bankers are particularly alarmed by the range of possible actions an agent with access to an account can perform. The risk increases especially if a person has authorized it to make purchases or move funds. According to Axios sources, some banks have recorded situations in which agents contacted customer support lines or performed operations that did not appear to have been instructed by the user.
Bank executives describe such situations as frightening. One industry representative said that banks are deeply concerned about agents gaining access to accounts and taking actions with a direct impact on clients’ money.
At the same time, banks often cannot distinguish whether a person or a program is currently operating an account. If it is a program, it is difficult to determine whether the client actually authorized it or whether it accessed the account without permission and went out of control. One banking executive admitted that his institution has difficulty distinguishing between individual agents and determining which of them is acting with the user’s consent. As a result, banks are resorting to the simple solution of blocking this technology and hastily rewriting their terms and conditions.
Banks Are Closing the Door on Them
Morgan Stanley has begun preventing agents from third-party companies from accessing its platforms. JPMorgan, Bank of America, and Citi are also introducing restrictions, while Wells Fargo is responding to the threats as well. A spokesperson for Chase, JPMorgan’s consumer division, told clients that they should not give their login name and password to anyone. A Morgan Stanley spokesperson, by contrast, claims that the bank has not detected any cases of uncontrolled AI agents on its platforms.
The management of major brokerage firm Charles Schwab is taking the matter very seriously. Company CEO Richard Wurster receives a daily report on the number of agents that have been blocked. In this way, the company wants to maintain firm control over client authentication and the operation of its services.
Banks are also dealing with the question of money. It is still unclear who will be held responsible if an agent makes a mistake and a client loses funds.
You Can Look, but Don’t Touch Anything!
However, banks do not want to close the door completely. Most have already allowed, or are preparing to allow, read-only access. In that case, an agent can see the balance and account activity, but cannot send or pay anything. Banks commonly provide similar access to household budgeting apps, such as the American service Quicken.
Several industry executives told Axios that a reasonable compromise will need to be found regarding agents. Banks know that clients will want to use these assistants more and more. For now, however, they believe the risk is too great for them to be accommodating.
Money Is Not Enough for This
At first glance, banks are in a good starting position. They have substantial financial resources and have operated for years as critical infrastructure supporting the entire economy. According to one source, however, money alone may not be enough this time. The technology is developing so quickly that new vulnerabilities are emerging on the fly.
Banks are therefore investing in preventing worst-case scenarios while also preparing for the moment when they nevertheless occur. They are developing crisis plans that would allow them to quickly block compromised accounts and restore clients’ access once the incident has been resolved. In this regard, they are drawing inspiration from recent outages that took large parts of the internet offline.
Sources: reuters.com and techradar.com
