Overview

A customer may share online banking credentials with an agent that then calls the bank with a complaint, without the bank having authorized that access. Meanwhile, Barclays plans to cut about 5,000 jobs and links efficiency changes to AI. Together, the reports put AI inside concrete decisions about account permissions and payroll.

Unapproved by a bank does not mean hacked

On October 23, CBS reported that U.S. banks were detecting third-party AI agents operating inside customers’ online accounts. In some cases, customers had given their usernames and passwords to agents the banks had not vetted. One example involved an agent calling a bank with a complaint before the account holder had approved it.

That distinction matters. CBS also made clear that the agents in its reporting were not breaking into accounts and draining funds. The dispute was more often about customers asking agents to act in ways their banks had not authorized. Unapproved automated access and stolen money are different risks.

White House AI adviser David Sacks suggested read-only access as one possible boundary: an agent could review transactions without being able to initiate them. The useful distinction is simple. Seeing information, proposing an action, and making that action happen should not be bundled into one permission granted by sharing a password.

Five thousand planned cuts put a price on AI investment

The same day, the Associated Press reported that Barclays planned to cut about 5,000 jobs, roughly 5% of its workforce. The bank linked the changes to organizational simplification and technology efficiency driven by AI; about 60% of the planned cuts would affect support functions, including developers.

That does not establish that AI directly replaces every affected worker. The reported figures describe the planned scale and where cuts would fall, not completed layoffs. What is clear is that companies are weighing AI investment against payroll in the same operating budget.

Both permissions and budgets need an audit trail

Together, the reports bring AI into two ordinary forms of organizational control. Banks need to know who authorized an action in an account; employers need to explain how technology changes staffing. A faster model cannot answer either question for them.

For customers, viewing and acting should be separate permissions. For workers, the savings attributed to AI should come with a clear account of whether they fund new investment, better service, or fewer jobs. The scale of bank-agent activity remains uncertain, and Barclays’ plan is not yet an outcome. Clear records of permission, responsibility, and numbers are more useful than saying AI is taking over everything.

Sources read for this article

CBS News, “AI agents are roaming around consumers’ bank accounts, and banks aren’t happy about it,” published October 23, 2026 at 1:46 PM EDT (timezone stated on page). https://www.cbsnews.com/news/ai-agents-customer-bank-accounts-security/

The Associated Press, “Barclays is latest to redirect spending toward AI as it trims its workforce,” published October 23, 2026 at 16:05:04 UTC. https://apnews.com/article/ai-layoffs-meta-microsoft-barclays-300bd6c2585b4a6996c2ac553d870055