Artículos Relacionados con Agent Security

The Use of Humans: Agentic Wallet and the Next Decade of Wallets

"Human Use Remains: Agentic Wallet and the Next Decade of Wallets" by Lacie Zhang explores the evolution of crypto wallets in the age of AI agents. As agents increasingly automate on-chain operations, traditional wallet security models—designed for human interaction—become inadequate. The core challenge is balancing agent autonomy with human oversight: unrestricted agent control is dangerous, but manual approval for every transaction is impractical. Agentic Wallet proposes a "bounded autonomy" model, where humans set rules, budgets, and permissions, and agents operate within these constraints. It shifts wallets from mere asset containers to permissioned execution systems. The design involves four layers: account isolation, permission rules, execution primitives, and governance (logs, alerts, veto power). Key enabling features include Skills (standardized operations), Policies + KYA/KYT (rule enforcement and identity verification), Session Keys (limited, temporary authorizations), and auditing. Current solutions (e.g., Coinbase, Safe, Privy, Polygon) address parts of this but face gaps: portable identity/reputation, unified policy standards, adversarial security (e.g., prompt injection), and cross-chain compatibility. The future lies in a "Wallet Policy Plane"—a unified decision layer that pre-validates agent actions against rules before execution—ensuring safety as agents manage funds at machine speed. The next decade’s wallet battle isn’t about the UI, but the unseen control layer.

marsbitHace 2 días 06:48