First regulated Swiss bank to use AI agents to test live on-chain transactions, with the client signing every action and private keys never leaving their device. Architecture is AI-model, and asset-agnostic, with the capability to turn client plain-text messages into multi-step on-chain operations for stablecoins, tokenized equities, gold and securities.
Sygnum, a global digital asset banking group, announced „it has become the first Swiss regulated bank to use an AI agent to test live digital asset market transactions, with the client retaining custody, consent and control at every step” – according to the press release.
In a first for Swiss banking, client-issued plain text instructions have been used to execute multi-step live transactions on a blockchain Mainnet. The Sygnum AI agent used these instructions and independently planned and prepared each step, reviewed the relevant smart contracts, and flagged potential transaction risks before presenting each transaction to the client for approval.
Transactions are only signed through the client’s self-custodial wallet on their own device, ensuring private keys do not leave client control at any time. Sygnum’s AI agent can be used to execute multi-step on-chain transactions for stablecoin transfers, asset swaps, on-chain lending positions, token wrapping and liquidity provisioning.
The pilot was built using a Model Context Protocol (MCP)* server built in-house by the AI@Sygnum team using Anthropic’s Claude as the underlying AI model.
The pilot transactions interacted with live mainnet smart contracts and market conditions using a sandbox-type approach. The pilot did not use Sygnum client data, wallets, accounts, funds or infrastructure, and used funds held in the individual’s own custody.
Thomas Frei, Head of AI and Data Analytics and AI@Sygnum lead at Sygnum Bank, said: “Connecting AI agents to wallets is foundational to where finance is heading. The next decade will see agents transacting, settling and interacting with markets on behalf of clients. The key challenge is doing this in a way that preserves – and even enhances – bank-grade consent, custody and trust. That is what we set out to solve, and what this pilot demonstrates: that a regulated Swiss bank can provide clients with the speed, convenience and accessibility of agent-driven execution, without ever giving up control of their assets.”
„Unlike architectures where AI agents with their own wallets transact autonomously, Sygnum’s approach enables clients to own, self-custody 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. This will benefit clients by enabling them to directly execute on-chain actions and access live market infrastructure, compose multi-step workflows without code and use their wallets natively.” – the bank explained.
This “human-in-the-loop” design is central to Sygnum’s integration of its agentic AI roadmap into all areas of the group’s global operations.
Sygnum’s MCP AI agent is not currently available to clients. Production deployment will be subject to full regulatory, compliance and security reviews and approvals. The pilot was conducted under controlled conditions. No Sygnum client CID, wallets or infrastructure were used.
__________
*MCP is a new open standard enabling AI and financial platforms to share context and data, streamlining transactions while maintaining client control and security safeguards. The MCP-based architecture is model- and asset-class agnostic, allowing the infrastructure to scale as the ecosystem develops.
The global blockchain AI market is projected to grow ~26.8% CAGR from USD 1.13 Bn in 2026 to USD 7.53 Bn in 2034.
Banking 4.0 – „how was the experience for you”
„To be honest I think that Sinaia, your conference, is much better then Davos.”
Many more interesting quotes in the video below:
