Financial institutions hesitate on MCP adoption due to KYC concerns

July 8, 2025 — The Model Context Protocol (MCP), launched in November 2024, has rapidly gained traction as a potential industry standard for AI agent interoperability. However, regulated sectors—particularly financial institutions—are holding back due to unresolved Know Your Customer (KYC) and compliance challenges.

Why Banks Are Wary

• Lack of KYC for Agents: Financial institutions require verifiable identity for all entities they interact with. MCP currently lacks a mechanism for agents to prove they represent licensed entities.
• Non-Deterministic Outcomes: Unlike traditional machine learning models, LLMs produce variable outputs, complicating risk assessments for banks.
• Audit Trail Gaps: MCP’s open-source nature means it’s still evolving, missing critical features like communication guardrails and traceability.

Industry Perspectives

• Sean Neville (Catena Labs): Compares MCP’s early stage to the pre-HTTPS web, emphasizing the need for foundational security standards.
• John Waldron (Elavon/U.S. Bank): Acknowledges MCP’s potential but highlights concerns about data traceability and risk leakage.
• Greg Jacobi (Salesforce): Notes that financial firms struggle with LLMs’ non-deterministic outputs, which clash with existing risk frameworks.

The Path Forward

While MCP and alternatives like Agent2Agent (A2A) and LOKA are under evaluation, financial institutions may delay adoption until protocols meet stringent compliance requirements.

For more insights, subscribe to VentureBeat’s newsletters.