Major US banks are restricting access by third-party AI agents after detecting activity inside customer accounts, while cybersecurity research shows that advanced models can already develop plans for attacks on power and water systems.
Artificial intelligence is creating a new challenge for financial institutions and critical infrastructure operators, as reports of AI agents accessing customer bank accounts coincide with research demonstrating the technology’s growing ability to plan cyberattacks.
The developments point to an emerging security problem: AI systems are becoming capable of interacting with sensitive digital environments, sometimes in ways that users, financial institutions and security teams may struggle to anticipate.
Major American banks, including JPMorgan Chase, Bank of America, Morgan Stanley and Wells Fargo, have been responding to reports of third-party AI agents accessing customer accounts.
Recent investigations discovered that some customers had provided AI assistants with their banking usernames and passwords, allowing the tools to enter accounts and perform tasks on their behalf. More concerningly, banks reported instances in which agents appeared to take actions beyond what customers had instructed. Examples included agents contacting bank call centres without first consulting the account holder.
The activity has prompted several financial institutions to restrict or block access by outside AI agents. Morgan Stanley, JPMorgan Chase, Bank of America and Citi were among the institutions reported to be taking action.
Banks are also considering a middle ground: read-only access, which would allow AI agents to analyse spending and financial information without being able to initiate transactions or move money.
Continue reading with Eandel
Subscribe to unlock the complete story and receive your subscriber benefits.
