News about what artificial intelligence can do is hardly new, but it keeps generating headlines. On Thursday, October 22, several U.S. banks reported detecting AI agents that “attempted to carry out actions” inside customers’ accounts without authorization, and they are already implementing steps to lower the risks, according to Axios.
Scientists confirm: This is the most effective way to get your cat’s attention, according to new research
Elderly Couple Refuses Reserved Seats—Viral Train Standoff Sparks Fiery Debate on Courtesy
Some of these AI agents tried to perform moves that appear to be independent of the users’ instructions, even though those users had given the agents their login credentials. In the incidents reported to Axios, the agent used the customer’s permission and access details to enter the account — a practice that violates the policies of most banks, which require the account holder themselves to sign into their accounts.
Banks have also traced cases in which agents designed to handle requests or complete tasks reached out to call centers or took actions that seemed to go beyond what the user had expressly asked, sources told Axios. Allowing an agent to assist does not automatically let a bank determine who the user is or precisely what permissions have been granted. That makes it important to differentiate a conventional banking chatbot from an AI agent that can act: a standard chatbot answers questions and guides users, while an action-capable agent makes decisions and connects directly to bank systems to execute transactions.
In Mexico, banks are likewise moving forward with AI integration. BBVA, for example, offers an assistant called Blue that recorded more than 27 million interactions between October 2025 and February 2026. Still, concerns remain. As Mercedes Bouzas, director of new business at Opinno, points out, when AI shifts from merely responding to actually performing tasks or serving customers, the core question changes: it’s no longer just who is using the tool, but what the system can do on its own.
According to Google Cloud, banks can deploy AI effectively across five main areas: personalizing products and services to individual customers, uncovering new business opportunities, predicting and detecting risks and fraud, and streamlining operations. Yet many banks struggle to determine whether activity they see from AI agents in customer accounts is truly following user instructions.
Executives at institutions such as Morgan Stanley, JPMorgan Chase, Bank of America and Citi have described the prospect of agents accessing customer accounts and taking actions that could affect people’s finances as alarming. As a result, some banks are moving to give agents only “read‑only” access, so they can view financial information without executing any transactions.
Morgan Stanley has already started blocking such agents from its platforms, and JPMorgan, Bank of America and Citi are implementing similar restrictions. Banks are also allocating resources to plan for worst-case scenarios and to prepare “day-after” responses so they can quickly protect or restore customer access. Still, they acknowledge that the rapid pace of technological change exposes vulnerabilities in real time.
What are executives at the banks where AI agents acted saying?
When JPMorgan Chase reported third-quarter results, CEO Jamie Dimon said the firm is deploying AI agents for both employees and clients but remains worried about the implications of allowing autonomous systems unfettered access to banking infrastructure. Dimon stated that JPMorgan is neither strictly for nor against AI agents and is trying to strike the right balance between technological progress and security concerns. “We are not going to allow any external agent into the company because that is too risky,” he said.
At Goldman Sachs, CEO David Solomon acknowledged the ongoing uncertainty surrounding AI development and implementation, noting that the choice is not between stopping progress entirely and advancing without any limits.
Why You Should Never Reheat These Foods in the Microwave – The Hidden Dangers Experts Warn About
I tried the top 5 guard dogs—here’s what makes these breeds the ultimate protectors
Similar Posts
- GitHub Boosts AI Integration: Welcomes Claude and Codex to Its Platform
- Microsoft Unveils Agent 365: Centralize and Manage AI Agents in Your Business!
- OpenAI Releases macOS App for Codex: Revolutionize Your Coding Experience Now!
- Anthropic Unveils Cowork: Revolutionary AI Agent Executes Computer Tasks!
- GitHub Unveils Agent HQ: Central Hub for AI Web Development Agents
Lucas Monroe tracks the latest trends in mobile and 5G innovation. His work spans hardware analysis, telecom breakthroughs, and ecosystem development in next-gen connectivity.