播客：保障 AI 智能体安全——身份、授权与 DPACT 框架
原标题：Podcast: Securing AI Agents: Identity, Authorization, and the DPACT Framework
AI 摘要
InfoQ 播客节目中，Sahil Agarwal 讨论了 AI 智能体在身份、授权与安全方面面临的挑战，并提出 DPACT 框架（委派、策略、可审计性、上下文、时间）作为构建受约束智能体系统的蓝图。他强调智能体应从简单的基于令牌的访问转向有边界的委派权限，并主张智能体应「代表」用户行事而非冒充用户，以防止权限提升。他还建议从清点资产开始逐步治理，并将有界任务授权作为未来智能体基础设施的一等特性。
正文节选
In this episode, Sahil Agarwal talks about the critical challenges of identity, authorization, and security in the age of AI agents. Sahil introduces the DPACT framework (Delegation, Policy, Auditability, Context, and Time) as a blueprint for building responsible, guardrailed agentic systems, moving away from simple token-based access toward bounded, delegated authority. Key Takeaways - AI agents are shifting from passive chat interfaces to unauthorised "delegated actors" that require robust sec