HONG KONG, April 21, 2026 /PRNewswire/ — On April 20, Ant Digital Technologies introduced its architectural vision for the agent economy at Hong Kong Web3 Festival — the “4R Full-Stack Architecture,” comprising four layers: Agentic Runtime, Payment Rails, Agent Registry, and Root Infrastructure — aimed at providing AI agents with foundational technical infrastructure covering identity, payments, risk control, and regulatory compliance.
In her keynote, Dr. Yan Ying, CTO of Ant Digital Technologies identified four fundamental fault lines in the current foundations of the agent economy: execution failures arising from prompt logic vulnerabilities, an accountability vacuum caused by AI’s lack of verifiable identity, transactional barriers stemming from payment gateways designed around human principals, and collaboration risks that emerge when unfamiliar agents cannot establish mutual trust. “This cannot be resolved by patching software,” she stated. “It requires a ground-up redesign at the infrastructure layer.”
The core product of the Agentic Runtime layer is DT Claw, which embeds the CARLI safety model to enforce behavioral constraints on agents at the execution level, supports multi-model compatibility and financial-grade compliance standards, and is designed to ensure that every AI operation is controllable, auditable, and recoverable.
The Payment Rails layer establishes a native on-chain payment channel that integrates agent-driven intelligent decision-making with verifiable credential chain technology, enabling precise identification of payment intent and end-to-end security while delivering full transaction transparency and immutability. For high-frequency micropayment scenarios, the platform builds a native instant settlement network supporting cross-chain, multi-asset seamless transfer and intelligent routing, significantly improving capital turnover efficiency. Additionally, by providing a standardized developer toolchain and a frictionless wallet integration experience, the solution substantially lowers both development barriers and end-user adoption costs — forming a payment closed-loop that balances financial-grade security with best-in-class usability.
The Agent Registry layer issues on-chain identities to each agent based on the DID (Decentralized Identifier) standard and ERC-8004, ensuring every instance of inter-agent collaboration is traceable and verifiable. The Root Infrastructure layer serves as the architectural foundation, leveraging Jovay Layer2 to achieve sub-120-millisecond transaction confirmation in support of AI micropayments, and combining ZKVM technology to enable off-chain computation with on-chain verification — resolving the computational trust problem inherent in the AI economy. As Yan Ying put it, “Root Infrastructure uses blockchain and privacy-preserving computation to provide agents with a tamper-proof contract execution environment. Even two agents with no prior relationship can establish trust through code and transact with confidence.”
AI is currently progressing from the Chat phase through the Action phase and into the era of the agent economy. Yan Ying argued that the defining transformation of this third phase lies not in AI becoming more intelligent, but in AI beginning to hold assets and exercise transactional authority. She noted that over the past decade-plus, Ant Digital Technologies has accumulated deep engineering expertise across financial-grade security, privacy computing, blockchain, and compliance systems — and that the 4R Architecture represents a ground-up research and development effort built upon that foundation.
