Sygnum Bank has conducted the first live trials of AI-agent-powered transactions in digital assets by a fully regulated Swiss banking institution. Announced this month, this pilot marks a pioneering step in blending artificial intelligence with blockchain operations while upholding stringent client safeguards and regulatory standards. The initiative stands out for its emphasis on client empowerment and security.
Users issue simple plain-text commands, which the AI agent interprets to orchestrate complex, multi-step operations directly on blockchain mainnets.
These can include stablecoin transfers, asset exchanges, on-chain lending adjustments, token wrapping, and liquidity management.
Crucially, the client must approve and sign every transaction through their personal self-custodial wallet on their own device.
Private keys remain entirely under client control at all times, eliminating any risk of the bank or AI holding custody.
This architecture relies on an in-house Model Context Protocol (MCP) server developed by Sygnum’s dedicated AI team.
Built around Anthropic’s Claude model, the MCP framework serves as an open standard that facilitates secure data sharing between AI systems and financial platforms.
Its design is intentionally flexible—independent of specific AI models or asset types—enabling handling of diverse instruments such as stablecoins, tokenized equities, gold, and securities.
This agnostic approach positions the technology for broad scalability as the digital asset ecosystem evolves.
Thomas Frei, Head of AI and Data Analytics and leader of the AI@Sygnum initiative, highlighted the broader vision: connecting intelligent agents to wallets represents the future direction of finance.
Over the coming decade, such agents are expected to handle transactions, settlements, and market interactions efficiently.
The primary challenge lies in maintaining consent, custody, trust mechanisms.
Sygnum’s pilot demonstrates how a regulated bank can deliver the speed and convenience of AI-driven execution while preserving full client ownership and control.
Unlike systems where AI agents operate with independent wallets and autonomous decision-making, Sygnum’s model keeps humans firmly in the loop.
AI serves as a powerful enhancer rather than a replacement for personal expertise and relationships.
Clients gain the ability to execute sophisticated on-chain strategies without writing code, while benefiting from always-available AI support that complements their dedicated banking advisors.
This human-centric philosophy aligns with four strategic pillars at Sygnum: elevating client experiences through AI-augmented relationships, driving regulated innovation for intuitive products, boosting operational efficiency by freeing staff for high-value tasks, and strengthening overall resilience, security, and governance.
The development arrives amid regulatory scrutiny of AI in finance. Authorities worldwide are highlighting risks such as agents exceeding user intent, prompting stricter rules on transparency, oversight, and accountability.
Sygnum has embedded governance frameworks covering data handling, risk assessment, and compliance to meet the highest bank-grade expectations. While the MCP AI agent is not yet available for general client use, the bank plans further regulatory, compliance, and security evaluations before any production rollout.