A Policy Algebra for Trust-Preserving Agentic AI Execution
作者: Bhaskar Tripathi, Anurag Kumar, Ramendra Kumar, Bhavesh Gadhe
分类: cs.AI
发布日期: 2026-08-17
备注: 7 figures, 10 tables, and 5 algorithms
💡 一句话要点
提出信任保护的代理AI执行策略代数以解决可靠性问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 代理AI 策略代数 信任保护 可靠性 企业智能 多代理系统 安全配置
📋 核心要点
- 现有的代理框架主要关注能力优化,缺乏对执行过程的可靠性保障,导致结果不可靠。
- 本文提出了一种策略代数,通过定义可靠性范围,确保代理在多种约束下完成任务。
- 实验结果显示,策略代数在94.8%的政策违规事件中进行干预,同时保持86.9%的任务完成率,显著提高审计完整性。
📝 摘要(中文)
基于大型语言模型的代理框架主要优化能力,即代理是否能够推理、检索信息、调用工具、委派工作并完成目标。然而,企业执行需要更强的可靠性属性。成功的结果如果是通过未经授权的数据访问、扩大委托权限、未批准的副作用、不可恢复的预算消耗或不完整的证据产生的,则不可靠。本文定义了可靠能力为路径属性:只有当代理在身份、配置、工具、数据、内存、预算、工件、批准和审计约束下完成任务时,才能被认为是可靠的。我们提出了一种策略代数,定义了代理能力可以被行使的可靠性范围。安全配置和运行时义务通过连接、交集、预算缩减、批准继承和证据积累进行组合,生成的组合既保护信任,又是满足所有治理输入的最不限制状态。
🔬 方法详解
问题定义:本文旨在解决现有代理框架在执行过程中的可靠性问题,现有方法未能有效控制数据访问、权限委托和预算消耗等方面的风险。
核心思路:提出了一种策略代数,通过定义可靠性范围,确保代理在执行任务时遵循身份、工具、数据等多重约束,从而实现信任保护。
技术框架:整体架构包括安全配置、运行时义务的组合,利用连接、交集等操作形成可靠性约束,并在多代理调用中传播限制。
关键创新:最重要的创新在于引入了成本感知的工件物化技术,能够在预算暴露增加时将开放式执行重定向为可恢复的结果。
关键设计:策略代数的设计包括对安全配置的动态调整、预算的实时监控和证据的累积,确保在执行过程中始终保持对信任的保护。
🖼️ 关键图片
📊 实验亮点
实验结果表明,策略代数在94.8%的政策违规事件中进行了有效干预,同时保持了86.9%的任务完成率,消除了配置单调性和零工件耗尽的违规情况,审计完整性提高至98.6%。
🎯 应用场景
该研究的潜在应用领域包括企业智能代理、自动化决策系统和安全敏感的AI应用。通过确保代理在执行任务时遵循严格的约束,可以提高系统的可靠性和信任度,进而推动AI技术在商业和社会中的广泛应用。
📄 摘要(原文)
Large language model-based agentic frameworks primarily optimize capability: whether an agent can reason, retrieve information, call tools, delegate work, and complete a goal. Enterprise execution requires a stronger property. A successful result is not reliable if it was produced through unauthorized data access, widened delegated authority, unapproved side effects, unrecoverable budget consumption, or incomplete evidence. This paper defines reliable capability as a path property: an agent is reliably capable only when it completes a task through action events that remain admissible under identity, profile, tool, data, memory, budget, artifact, approval, and audit constraints. We propose a policy algebra that defines the reliability envelope within which agent capability may be exercised. Security profiles and runtime obligations compose through joins, intersections, budget narrowing, approval inheritance, and evidence accumulation; the resulting composition is both trust-preserving and the least restrictive state satisfying all governing inputs. The algebra also propagates restrictions across multi-agent calls and introduces cost-aware artifact materialization, which redirects open-ended execution toward a recoverable outcome as budget exposure grows. The evaluation is interpreted as a reliability-capability trade-off rather than a capability benchmark: the policy-algebra runtime intervenes on 94.8% of policy-violating events while retaining an 86.9% task-completion rate, eliminates the observed profile-monotonicity and zero-artifact-exhaustion violations, and increases audit completeness to 98.6%. The method provides researchers and practitioners with formal correctness conditions, executable decision semantics, and trace evidence for building agents that are not only capable, but reliably capable.