Swiss bank Sygnum has achieved a significant milestone by becoming the first regulated Swiss bank to complete live AI-driven digital asset transactions with client retaining custody, consent and control at every step. According to the latest announcement from Sygnum, the bank ran live on-chain transactions through an AI agent while clients maintained full custody of their assets. The pilot utilizes the bank's in-house Model Context Protocol (MCP) server, powered by Anthropic's Claude as the underlying AI model. As reported by Thomas Frei, Head of AI and Data Analytics and AI@Sygnum lead at Sygnum Bank, connecting AI agents to wallets is foundational to where finance is heading, with the next decade expected to see agents transacting, settling, and interacting with markets on behalf of clients.
The system ensures private keys never leave the client's device throughout the entire transaction process. As reported by Sygnum, clients submit plain-text instructions, and the AI agent independently plans each step, reviews smart contracts, and flags potential transaction risks before presenting each transaction to the client for approval. Every signature occurs on the client's own device through a self-custodial wallet, maintaining complete client control over their assets. According to Thomas Frei, unlike architectures where AI agents with their own wallets transact autonomously, Sygnum's approach enables clients to own, hold and fully control their own wallets and assets at all times. This reflects a core principle of the AI@Sygnum program, where AI augments and enhances, but does not replace, human decision making or trusted personal relationships.
The AI agent handles various use cases including stablecoin transfers, asset swaps, on-chain lending, token wrapping, and liquidity provisioning. According to Sygnum, this represents a significant advancement from traditional AI advisory tools to active transaction execution capabilities. The bank emphasizes that connecting AI agents to wallets is foundational to the future of finance, with the next decade expected to see agents transacting, settling, and interacting with markets on behalf of clients. The MCP-based architecture is model- and asset-class agnostic, allowing the infrastructure to scale as the ecosystem develops.
Production deployment of Sygnum's AI agent is subject to full regulatory, compliance and security reviews and approvals. As reported by Sygnum, the bank's AI governance on data, risk management, transparency and accountability is designed to fully meet the highest regulatory, bank-grade standards. The pilot was conducted under controlled conditions with no Sygnum client CIDs, wallets or infrastructure used. The four key pillars of Sygnum's organisational AI strategy include client experience, regulated innovation, organisational efficiency, and operational excellence. The MCP AI agent is not currently available to clients, reflecting the bank's commitment to ensuring proper regulatory frameworks are in place before broader deployment. Commercialization depends on resolving outstanding regulatory and security questions, as Swiss financial regulators, including FINMA, have yet to issue specific guidance on AI-driven trading within banking frameworks.
