The open-source protocol 'Model Context Protocol (MCP)' supporting interaction between AI agents is spreading rapidly, but industries under regulation are still maintaining a cautious attitude. Particularly, financial institutions are hesitant to participate due to MCP not meeting Know Your Customer (KYC) requirements.
MCP has been showing a steep adoption rate since its release as early as last November, potentially becoming an industry standard. However, heavily regulated industries like banks and medical institutions continue to maintain a conservative stance. These industries were early investors and experimental companies in AI utilization, but are currently focusing more on risk assessment than actual implementation of new protocols like open agent exchange systems.
The MCP system, where AI agents communicate in real-time and access various data, certainly provides advantages in efficiency and innovation. However, financial institutions handling high-risk areas such as loans, investments, and customer information analysis still prioritize **security** and **compliance**. If revealed technical vulnerabilities could lead to potential **data exposure** or **audit deficiencies**, such a system cannot be adopted.
John Waldron, Senior Vice President of Elavon, a US bank subsidiary, stated, "We are reviewing various possibilities for MCP connections using AI agents, but there are risks we cannot definitively guarantee until clear standards emerge." He emphasized that it is necessary to review whether MCP messages merely mediate communication or potentially expose sensitive information.
Greg Jacobi, Banking Industry Strategy Lead at Salesforce, pointed out that while existing machine learning models were reproducible and fit well into risk assessment frameworks, large language models (LLMs) cause significant conflicts due to their probabilistic nature. Consequently, the financial sector is recognizing the need for LLMs to be **managed differently** from existing models.
Additionally, when external agents access a system, it is crucial to identify whether the request originates from an authenticated organization. However, neither MCP nor Google's Agent-to-Agent (A2A) protocol has completely resolved this issue. Sean Neville, co-founder of Catena Labs, emphasized that to satisfy existing KYC regulations, agents must clearly disclose their identity, risks, and delegation information, which will be core to future standard protocols.
Currently, some financial institutions are testing limited environment implementations or considering alternative communication methods like A2A, but **legal and technical foundations are still required** to fully enter the open agent era. Despite MCP emerging as a **common language** across the AI industry, highly regulated sectors continue to contemplate how such technology can be utilized.
For real-time news...Go to Token Post Telegram
<Copyright ⓒ TokenPost, Unauthorized Reproduction and Redistribution Prohibited>
