Skip to content

Translation in progress Some entries still include text in the original language.

Failure-First·Indexed by this site · Originally published

Foundation Protocol:为人类—AI 社会构建安全基础设施

[Daily Paper] Foundation Protocol: Building the Safety Infrastructure for a Human–AI Society

AI summary

Foundation Protocol(FP)论文提出把智能体交互的协议层当作安全边界,为跨协议智能体行为建立可核验的问责基础。FP 认为 MCP、A2A、A2UI 各自只覆盖单一边界,一条工作流同时跨越多重边界时溯源会断裂、问责无法大规模核验;它把协调层分为实体与信任、传输与路由、交互与组织、监管与监督四个平面,将策略执行与溯源钩子嵌入通信底座。论文称这与 HarnessAudit(2605.14271)的实证相符:多智能体 harness 的安全违规集中在资源访问与智能体间信息传递。

Read originalfailurefirst.org