Banks Require Revocable AI Identities and Ongoing Trust Frameworks

Financial institutions are accelerating the deployment of artificial intelligence agents capable of initiating payments, approving transactions, and managing account freezes. However, these advancements highlight a critical challenge: traditional authentication frameworks are predicated on the assumption that a human is always involved. As the adoption of agentic AI increases, banks must address an emerging authentication crisis that necessitates innovative controls.

In a recent statement, David Barnhardt, a strategic advisor specializing in fraud and anti-money laundering at Datos Insights, emphasized the need for banks to obtain visibility into both the intent and integrity of AI systems, especially when agents diverge from their predefined behaviors. Financial institutions must transcend standard identity checks, instead validating delegated authority through managed, revocable digital identities specifically designed for AI applications.

Barnhardt elaborated that the evolution of authentication is imperative; it should no longer focus solely on verifying identity but also on validating and confirming delegated authority. To establish an effective framework, he advocates for managed digital identities that incorporate cryptographic proof, ensuring that all actions can be traced to the human who authorized them and the AI agent performing them.

The implications of this shift extend beyond basic authentication measures. As AI agents gain more autonomy, the nature of fraud is anticipated to evolve. Barnhardt warns that threat actors may transition from stealing credentials to compromising AI agents, compelling banks to rethink their strategies for authentication and authorization on a larger scale. To mitigate this risk, institutions need to be equipped to revoke certificates, disable compromised agents, and employ parallel models that can distinguish standard agent behavior from fraudulent activity.

In an interview with Information Security Media Group, Barnhardt discussed weaknesses in static authentication controls and the challenges of verifying intent and integrity as agents evolve. He underscored the importance of proactive preparation for emerging fraud risks and regulatory demands.

Barnhardt brings over 22 years of experience in fraud mitigation within the banking and payments sectors. His background includes positions at the Bank of America, Early Warning, Giact, and Conduent, where he developed patents currently utilized to combat fraud and reduce payment risks.