Unvetted third-party AI agents access bank accounts

Major US banks are blocking or restricting third-party AI agents from accessing bank accounts. Indian banks' preparedness remains unclear
AI bots looking at computers
Mohamed Nohassi/Unsplash
Published on: 

In a serious security breach, major US banks have detected unvetted third-party AI agents operating inside customer accounts, with some agents attempting to take actions that appear to be independent of user instructions, Axios reported on October 22.

The danger arises when customers allow a third-party AI agent access to their account's login credentials and grant it purchase authority or the ability to transfer funds. Some banks have also tracked AI agents contacting call centres or appearing to take actions without explicit directions from the customer.

One bank executive told Axios they struggle to ascertain whether an agent has been authorised by a user or has gone rogue.

Industry insiders have described the situation as "horrifying".

Banks including Morgan Stanley, Bank of America, JPMorgan, and Wells Fargo are rushing to contain the threat. According to Reuters, Morgan Stanley has blocked third-party AI from accessing its platforms, though it has not seen any instances of unauthorised or rogue AI agents.

In a statement to Retail Banker International, Citi said: "To help safeguard customer accounts and information, we do not permit third-party AI agents or assistants to log into a customer's account or transact directly within our digital banking platforms."

Customers can, however, continue to authorise limited, read-only access to their account information with certain Citi-approved open banking providers.

Citi

The India landscape

"Indian banks are increasingly focusing on deployment of AI-driven chatbots and virtual assistants to enhance customer experience and provide personalised service," says the Reserve Bank of India (RBI). "These technologies provide customers with instant responses to queries, streamline account management, and offer personalised financial advice."

The RBI has issued detailed guidelines for banks deploying AI in their systems. However, the banks' policies and preparedness to deal with the risks posed by third-party AI agents used by customers are unclear.

The Financial Standard sent a query to HDFC Bank and Axis Bank, which have made significant investments in their in-house AI assistants, about their approach to the emerging threat. The story will be updated when they respond.

For now, the responsibility to protect their account appears to fall largely on the customer: the RBI classifies any financial loss arising from a customer sharing their password, PIN, OTP, etc with a third party as customer negligence. The customer has to bear the loss till they report it to their bank.

A report by consulting firm PwC says customers must make a voluntary trade-off between greater privacy and customised service.

Those that highly value their privacy will be giving fewer data points to the AI models, resulting in less customised (and more general) service. Comparatively, customers that consent to the sharing of private data will enjoy a more personalised service experience.

PwC

Ultimately, dealing with the AI challenge isn't just a question of technology.

"We face an unusual technological contest in which both the defender and the attacker increasingly have access to similar tools," RBI deputy governor Rohit Jain said at the SBI Banking & Economic Conclave in September. "The advantage will not necessarily belong to the institution that deploys the most sophisticated technology. It may belong to the institution that can understand, govern and deploy technology with the greatest degree of discipline and foresight."