Charles Schwab is shutting out outside artificial intelligence agents that try to move money from client accounts, and its chief executive says the company has no clear answer yet on who pays if one of those agents makes a costly mistake.
Rick Wurster, Schwab’s CEO, told Bloomberg Television on Friday that the firm is blocking unauthorized agents, including Meta’s Muse. The core problem, he said, is what happens when a client hands over login credentials to an agent that then acts for them. “We can’t control it, the client can’t control it, something could go wrong and who’s liable?” Wurster said.
That question has no settled answer in the United States, where lawmakers in Congress are still debating how liability should work for AI agents.
Wurster said he receives a weekly report on the blocked attempts. The large agents are stopped in the tens and twenties each week, he said. In his view, the numbers show that clients are not yet ready to give AI tools access to their accounts.
Muse is Meta’s personal AI agent. It can book, buy and negotiate on behalf of its users. Schwab is not trying to keep outside technology out altogether. Instead, the firm plans to set up model context protocol (MCP) servers, which are secure links that let a client’s chosen AI model connect to Schwab’s systems under rules the company sets. Through those links, outside tools will be able to check balances and run research. They will not be able to move money, Wurster said.
He also wants customers to use Charley, Schwab’s own AI assistant, which the company announced on Sept. 30.
Schwab is not acting alone. On Tuesday, JPMorgan Chase CEO Jamie Dimon said the bank would not let outside AI agents act inside its systems. Speaking on the bank’s earnings call, Dimon warned of “rogue agents,” according to a report from Cybernews.
JPMorgan already allows some third-party tools to view account data, but those tools cannot send money or pay bills, the same report said.
The moves come as regulators in the United States and Europe press banks on how they use AI agents. Both firms are drawing a line between letting software look at financial information and letting it take action with customer funds.
The security concerns are not limited to banks. MCP links have had their own security problems, and Google and JPMorgan are among the companies that have fixed the same flaw.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.