Daiwa Securities Tests AI Agents for Blockchain Token Swaps

Daiwa Securities Group, alongside its subsidiaries, successfully tested AI agents capable of executing blockchain-based token swaps within a controlled environment on Friday.

The trial, reported by IBTimes JP, mandates human approval for all transactions, ensuring a verifiable audit trail from initial natural language instructions to final execution.

The experiment involved Daiwa Securities and the Daiwa Institute of Research utilizing a test stablecoin and a test security token. By bypassing proprietary, single-vendor silos, the team opted for open standards to manage agent coordination and post-trade verification.

The system architecture functions as a gated execution layer. An AI agent parses a user’s natural language intent, but the actual on-chain transaction remains locked behind a human-in-the-loop requirement. Every step—from the initial request to the final settlement—is logged in a linked record, allowing for systematic review and compliance auditing.

The technical core of the trial focused on the integration of disparate systems. By linking the instruction, the approval, and the transaction result, the researchers created a transparent pipeline. This is a departure from “black box” automated trading, where the decision-making process is often obscured by the complexity of the underlying smart contract logic.

The Daiwa Institute of Research designed the environment specifically to stress-test these audit logs. By restricting the agents to a predefined set of trade types, the developers could isolate the technical hurdles associated with cross-exchange token swaps.

The trial serves as a pragmatic look at the limits of autonomous finance. Daiwa’s strategy highlights the tension between high-frequency execution and the absolute requirement for legal compliance.

The group is now shifting its focus toward identifying which specific tasks are ripe for delegation. This involves evaluating how these agents will eventually interact with external systems and third-party AI models. The challenge lies not just in the code, but in the governance frameworks required to manage risk in an environment where agents, not just humans, initiate trades.

The trial highlights three primary areas of focus for the organization’s future development:

  • Auditability: Maintaining a tamper-proof record of agent reasoning and human confirmation.
  • Interoperability: Moving beyond single-vendor constraints to utilize open, standardized protocols for asset movement.
  • Governance: Establishing the legal and compliance boundaries for autonomous, cross-platform financial actions.

While the current trial is confined to test assets, the underlying infrastructure is designed to scale with the tokenization of real-world securities. The companies have not yet decided to launch a commercial product based on these trials, focusing instead on the foundational requirements for future financial services.

Daiwa Securities Group traces its current structure to 1999, when it was established as a joint venture with Sumitomo Bank under the name Daiwa Securities SB Capital Markets.

The group later dissolved that partnership in 2010 after Sumitomo Mitsui Banking Corporation acquired rival Nikko Cordial, reacquiring full ownership and absorbing the entity into Daiwa Securities Co. Ltd. Today it stands as Japan’s second-largest securities brokerage after Nomura Securities, with an international investment banking arm, Daiwa Capital Markets, operating across Asia, Europe, and North America.

As the firm evaluates the results, the next phase will involve “assessing links to external systems and third-party AI agents, alongside further legal, compliance and governance work,” according to the companies.

Photo of author

Technology Desk

Technology Desk at Archyde.

Newsian: Dow Rises 0.92% as Nasdaq Falls 0.78% on AI Selloff