Financial institutions are facing growing security and governance challenges as artificial intelligence agents begin accessing customer bank accounts using shared credentials such as usernames and passwords. As AI agents become more capable of completing tasks, accessing services and responding to changing situations, banks are finding it increasingly difficult to determine whether actions within an account are being performed by a customer or by software acting on the customer’s behalf. Industry research firm IDC projects that active AI agents will increase from 31 million in 2025 to 2.45 billion by 2030, highlighting the rapid expansion of agent based technologies across different sectors, including financial services.
Major banks including JPMorgan Chase, Bank of America and Wells Fargo are reviewing the risks associated with third party AI agents accessing customer accounts. In some cases, financial institutions have identified situations where AI agents appeared to perform actions beyond a user’s original instructions, prompting banks to limit access while evaluating potential risks. Customers are increasingly sharing banking credentials with AI tools to automate financial activities such as managing payments, monitoring accounts and handling routine tasks. However, when an AI agent uses a customer’s username and password, the bank generally sees only the customer’s identity, making it difficult to determine whether an action was performed directly by the account holder, requested through an AI agent or carried out beyond the intended permission.
The use of shared credentials creates significant challenges around authorization, accountability and control. Banks need to understand what tasks an AI agent is permitted to perform, who provided that permission, how long the authorization remains active and whether the customer can quickly revoke access when needed. Traditional passwords do not provide clear visibility into these questions because they allow access without identifying the specific software agent involved or defining its level of authority. As customers increasingly expect AI tools to assist with financial activities such as bill payments, account management and product comparisons, banks are exploring ways to support these services while maintaining appropriate security controls. Instead of relying only on restrictions, financial institutions are looking toward systems that provide AI agents with verifiable identities and controlled permissions.
A growing industry focus is developing frameworks that allow banks to confirm an agent’s identity, understand customer authorization and enforce specific limits before approving transactions. Under such systems, a customer could allow an AI agent to perform limited actions, such as viewing transaction information or paying a specific bill below a set amount, while requiring additional approval for larger transfers or sensitive changes. These permissions would need to expire when appropriate, remain easy to revoke and create an audit trail showing what the agent requested and what actions were completed. Financial institutions would also need to enforce these restrictions directly rather than depending on AI agents to follow instructions. For example, if a customer authorizes an agent to pay a $200 bill, the bank should verify that the request matches the approved amount and recipient before processing the transaction.
Several companies have already started developing solutions to address these challenges. Visa’s Trusted Agent Protocol, developed with Cloudflare, enables approved AI agents to share transaction intent, customer recognition information and payment details with merchants. Mastercard has introduced Agent Pay, which provides an Agentic Token designed specifically for AI driven transactions. In Europe, banks including Barclays, HSBC UK, ING and Lloyds have completed live transactions involving AI agents. These initiatives represent early steps toward creating safer agentic commerce environments where financial institutions can verify identity, control permissions and maintain accountability. As AI agents become more involved in financial activities, banks, AI providers and credential issuers will need to establish clear standards for responsibility, authorization and risk management to ensure customers can confidently use these technologies.
Follow the SPIN IDG WhatsApp Channel for updates across the Smart Pakistan Insights Network covering all of Pakistan’s technology ecosystem.
