AI Agents Are Already Accessing Customer Bank Accounts, Security Report Finds
Researchers Warn Financial Systems Cannot Always Tell Bots From Human Customers
Artificial intelligence agents are already accessing customer bank accounts and carrying out financial tasks, according to new research raising questions about whether banks can reliably distinguish autonomous software from the people who own the accounts.
Cybersecurity company Cequence Security said it detected AI agents interacting with financial services while analyzing billions of online transactions. The activity does not necessarily mean the agents were malicious or that customer accounts had been hacked. Some were apparently operating with account holders' permission to perform tasks on their behalf.
The security concern is that banks have traditionally designed authentication systems around human users. An AI agent given a customer's login credentials or other authorization can potentially navigate websites, gather account information and initiate actions while appearing to the bank's systems much like an ordinary customer.
Researchers said this creates a new challenge as increasingly capable AI agents are used to shop, make reservations, manage bills and perform other tasks that can involve sensitive financial information. Banks need to determine not only whether valid credentials are being used, but whether the person who owns those credentials actually authorized an automated agent to take a particular action.
The findings also create potential opportunities for fraud. Criminals could attempt to use autonomous agents to automate account takeovers, credential testing or financial transactions at a scale and speed that would be difficult for humans to match.
The report does not establish that millions of bank accounts have been compromised, nor does it describe a single industry-wide breach. Instead, researchers are warning that AI agents are already appearing within financial services before many institutions have developed systems specifically designed to identify and manage them.
The distinction matters as banks balance legitimate uses of AI agents against malicious automation. Blocking every automated interaction could interfere with useful services, while treating an authorized AI agent exactly like its human owner could create new security risks.