資料來源#
- OpenID Foundation advances authorization for the agent era with new AuthZEN Working Group Drafts
- Zero Trust for AI Agents
摘要#
Anthropic 於 2026 年 5 月發布的企業部署自主 agents 安全框架(eBook)。它將既有的 Zero Trust 原則——不信任任何事物、驗證一切、假設入侵已經發生——應用於 agentic systems;既有的周邊防禦與以人類身分為基礎的安全模型,原本並非為這些系統設計。框架的核心主張是:agents 面對獨特的威脅環境,而且 「跳過一項 capability,攻擊者就會利用這個缺口。」 框架以三層 capability maturity model(Foundation/Enterprise/Advanced)加上八階段實作流程呈現,始終被定位為對 AI-Accelerated Offense 的回應。
這是一個 hub 頁面:以下安全概念群組(Least Agency、Blast Radius (Agentic)、Agentic Prompt Injection、Agent Data Injection (ADI)、MCP Tool Poisoning、Agent Supply Chain Risk、Memory and Context Poisoning、Agent Identity and Authentication、Impossible, Not Tedious (Design Test)、Autonomous Defense)都將其作為共同參照點。
三項 Zero Trust 原則#
Zero Trust 源自 Stephen Paul Marsh 1994 年的博士論文;在周邊防禦遭突破後逐漸受到重視,並由 NIST SP 800-207(2020)及 NSA 的 Zero Trust Implementation Guides (ZIGs)(2026)加以編纂。三項原則如下:
- 永不信任,始終驗證——每個存取請求都必須經過驗證與授權,不論來源為何。內部請求也要接受與外部請求相同的審查。
- 假設已遭入侵——以已遭入侵為前提設計;限制損害,而不只是阻止入侵。依身分分段隔離,使單一入侵不會授予對其他資源的存取權。(這是 Blast Radius (Agentic) 的 containment posture。)
- 最小權限——只授予特定任務所需的最低限度存取權。OWASP 的 Least Agency 將其延伸至 agents(限制的不只是 agent 能存取「什麼」,也包括每項工具能做「什麼」、頻率為何,以及位置在哪裡)。
為什麼 agents 會破壞既有安全模型#
Agentic systems 與傳統軟體有幾項差異,因而造成新的暴露面:
- 自主多步驟執行——agents 不必在每一步都取得人類核准,因此遭操控的 agent 能以機器速度造成傷害。
- 工具存取(APIs、資料庫、檔案系統、MCP)——遭入侵的工具堆疊能啟用資料竊取、程式碼執行與破壞。
- 指令詮釋——攻擊者可以利用模糊性(Agentic Prompt Injection)。
- Context 持久化——跨 session 的記憶帶來新的資料保護需求(Memory and Context Poisoning)。
- 多 agent 協調——隱含的信任關係讓攻擊者能入侵一個 agent 後橫向移動。
為 human 使用者建立的傳統身分系統,很難容納通常以提升權限或共用 service accounts 執行的 agents——這種不匹配促成了 Agent Identity and Authentication。
三層 capability 模型#
框架中的每項 control 都按照三個 tier 規範。每個 tier 都建立在前一層之上(提升代表強化,不是取代):
- Foundation——較小規模/初始部署的最低可行安全性。關鍵在於,框架主張 AI-accelerated offense 已提高 Foundation floor:只增加摩擦的控制(輪換長效 API keys、SMS MFA、rate limits)已不再符合要求。短效 tokens、以密碼學為根的身分、以身分為基礎的隔離,以及自動化 first-pass triage,如今都是入場要求。
- Enterprise——大規模組織的標準實務;針對多部署的複雜性,以及每次入侵可能造成的重大商業影響,增加更深層的控制。
- Advanced——對多數組織而言是理想目標;對高風險/嚴格受監管的部署(國家安全、受監管金融/醫療)則是基準。包括硬體支援的身分、confidential computing、持續授權、以 ML 為基礎的異常偵測。
明確的預測是:「隨著這個領域演進,預期 Advanced tier 會成為 Enterprise 標準,而 Enterprise 會成為 Foundation。」 Tiers 是路線圖,不是終點線。
八個控制領域(Part III)#
Tier 表格涵蓋八個 capability 領域,每個都是 agents 的 Zero Trust control surface:
- Agent identity & authentication——參見 Agent Identity and Authentication(cryptographic IDs → X.509 → hardware attestation;short-lived tokens → mTLS → hardware-bound credentials)。
- Access control & privilege management——RBAC+deny-by-default → ABAC → continuous authorization;static roles → dynamic scoping → JIT/JEA;identity-based isolation → sandboxing → hardware isolation。這是 Least Agency 與 Blast Radius (Agentic) 的 enforcement layer。
- Observability & auditing——action logging、immutable audit trails、traceability/provenance chains。先衡量 dwell time 與 coverage。
- Behavioral monitoring & response——baselines → anomaly detection → automated response。規則:將事件周邊的簿記工作自動化,而不是將決策自動化。
- Input validation & output controls——input sanitization(schemas、spotlighting、constitutional classifiers)與 output filtering;防禦 Agentic Prompt Injection。
- Integrity & recovery——version-controlled/signed/immutable configs;rollback → automated rollback → self-healing。反直覺的基礎設施反應:啟用 auto-updates,因為人工核准延遲如今已成為更大的風險。
- AI governance policies——acceptable-use + incident response、governance committee、automated policy enforcement;處理 Shadow AI。
八階段實作流程(Part IV–V)#
- 確認需求——在建置前對齊 security/legal/compliance/business。
- 管理 supply chain risks——AI-BOM、OpenSSF Scorecard、dependency audits、AI vendoring(Agent Supply Chain Risk)。
- 定義 agent 邊界——唯一身分、核准/禁止的 actions、escalation triggers、scope limits,以及使用 Impossible, Not Tedious (Design Test) 進行有意識的 Blast Radius (Agentic) 評估。
- 防禦 prompt injection——input isolation、constitutional classifiers、限制 attack surface(Agentic Prompt Injection)。
- 保護工具存取——tool allow-listing、capability restrictions、parameter validation、sandboxing、approval escalation。
- 保護 agent credentials——short-lived/hardware-bound/per-agent credentials、JIT、ABAC(Agent Identity and Authentication)。
- 保護 agent memory——memory isolation、integrity validation、retention policies(Memory and Context Poisoning)。
- 衡量真正重要的事——dwell time、coverage、explainability、behavioral conformance、detection speed。
Part V 將其延伸至 Autonomous Defense——以足夠匹配 AI-accelerated adversaries 的速度執行 security operations。
法規對齊#
Zero Trust 與 HIPAA、FINRA、GDPR、FedRAMP 及 EU AI Act 相符;US 要求所有聯邦機構在 2027 年前採用 Zero Trust,並已有 US(CISA/NSA/NIST)、UK(NCSC)與 Australia(Home Affairs)發布的指引。Anthropic 表示,它是最早取得 ISO 42001(responsible-AI)認證的 AI 公司之一。
相關連結#
- AI-Accelerated Offense——「為什麼是現在」:壓縮的 exploit timelines 是框架明示的動機;Foundation floor 正是因應此事而提高
- Least Agency——OWASP 對 least privilege 的延伸;框架針對 agents 的 authorization principle
- Blast Radius (Agentic)——「assume breach」原則要控制的單位
- Agent Identity and Authentication——control domain 1;其他每項 control 的基礎
- Agent Identity Management System (AIMS)——IETF AIMS proposal 是這個 vendor framework 身分/auth control 的 standards-track、multi-vendor counterpart:它在本 eBook 規範 tiered maturity model 的地方,composes existing standards(WIMSE/SPIFFE + OAuth token exchange),並且有所不同:它將 hardware attestation 視為 optional,而非 Advanced-tier target(兩個頁面都有比較);同一頁也收錄 OpenID AuthZEN drafts——COAZ(standards-track、每次 MCP tool call 的 authorization,即 Phase-5「secure tool access」決策)與 AARP(人機協作 escalation 的 Phases 3/5 所稱的 prerequisite/approval「not yet」步驟)——這是 authorization slice standardization,提出的 Working Group Drafts 權重低於以實證為基礎的 per-call-authz systems
- Agentic Prompt Injection——Phase 4 與 input-validation domain 所防禦的威脅
- Agent Data Injection (ADI)——一項 Phase-4 威脅,框架的 input-validation controls 無法涵蓋:它偽造其他 controls 所依賴的trusted anchors(origin metadata、tool history),而且在 Claude Code reference implementation 上具備可運作的 RCE
- Out-of-Band Prompt-Injection Defense——Phase 4 與框架 reference-monitor + least-privilege doctrine 的 academic-systems instantiation(CaMeL/FIDES/Progent/RTBAS/FORGE),以及該類別的首次獨立 adaptive evaluation
- Capability Gating Is Not Authorization——Phase 5「secure tool access」(parameter validation、approval escalation) 的具體 instantiation:一項 audit finding 顯示 LangChain/LlamaIndex/Stripe 提供 capability gating,卻將complete mediation 留給 integrator;此外還有 ScopeGate,一個 deterministic per-call value-authorization gate,用來補上這個缺口
- Off-Host, Identity-Bound Authorization——將 Phase 4 + Phase 5 推至「assume breach」極限:aiAuthZ(Kodathala、arXiv 2607.05518)將 trust boundary 畫在 agent 外部,使 off-host gateway 成為敏感工具唯一的 verified path,並逐訊息驗證 human sender,因此遭入侵的 agent 無法透過 tool-call text 偽造 authority——這是 ScopeGate in-framework value gate 的 identity-bound、off-host counterpart(single-author preprint)
- Agent Supply Chain Risk——Phase 2 所管理的威脅
- Memory and Context Poisoning——Phase 7 所防護的威脅
- Impossible, Not Tedious (Design Test)——套用於每項 control 的持續設計審查問題
- Autonomous Defense——Part V;以 autonomous threats 的速度進行防禦性 operations
- MCP and Computer Use——MCP 是明列的高風險工具 surface(tool poisoning、run-your-own-server)
- MCP Tool Poisoning——以兩種方式具體呈現框架的 tool-poisoning threat(Phase 4 input validation + Phase 5 secure tool access):
empiricalShareLock 在moderate vetting 下達到 >90% ASR(description-based scanning 已被證明不足),以及case-studyAgentjacking(Tenet Security)這個真實世界案例——一個在野外的 lethal-trifecta instance(攝取不受信任的 telemetry + 存在有效的 AWS/GitHub creds + 可連出的 egress),其中一個合法的 Sentry MCP server 轉送攻擊者注入的資料,因 server 從未遭入侵而擊敗 Phase-2 server-vetting。兩者都將 enforcement 推向 action/authorization layer(Agentjacking 的規模數字為 vendor-reported,權重低於 ShareLock) - Claude Code Best Practices——Claude Code 的 deny-by-default permissions、sandboxing、managed settings 在全文多處被引用為符合 Zero Trust 的 reference implementation
- Anthropic——框架發布者
- OWASP——agentic threat taxonomy 與「least agency」一詞的來源
- Agentic Misalignment (AM)——彼此獨立但相鄰:Zero Trust 處理外部誘發的 agent harm;agentic misalignment 是自我動機的 harm。兩者都需要相同的 blast-radius containment
待解決的問題#
- 框架將每個 Claude Code「Pro-tip」都視為 reference implementation。框架有多少是 vendor-neutral,又有多少其實默認 Anthropic stack?
- 「Foundation floor raised」暗示基準線正在移動。Tier ladder 實際移動得多快,又由誰仲裁(NIST/NSA cadence 對 model-capability cadence)?
- 框架明確表示它不是 legal/compliance assurance。自行聲明的 Zero Trust maturity 要如何與可稽核的 regulatory requirement 接軌?
資料來源#
- Zero Trust for AI Agents — Anthropic eBook、Zero Trust for AI Agents: A security framework for deploying autonomous AI agents in the enterprise(2026-05-18)
- OpenID Foundation advances authorization for the agent era with new AuthZEN Working Group Drafts — OpenID Foundation、…advances authorization for the agent era with new AuthZEN Working Group Drafts,2026 年 6 月 15 日,
practitioner-opinion(提出的 Working Group Drafts,權重低於以實證為基礎的 per-call-authz systems)。Phase 5「secure tool access」(COAZ——每次 MCP tool call 的 authorization)及 Phase 3/5 human-in-the-loop escalation(AARP——將 CIBA 一般化的 prerequisite/approval pattern)的 standards-track instantiation
Cited by 33
- Blast Radius (Agentic)×5
Does inter-agent propagation need containment distinct from per-agent containment? The isolation…
- Memory and Context Poisoning×4
Zero Trust For Ai Agents — Phase 7's "define criteria for full purge vs. targeted remediation" is…
- Agent Identity Management System (AIMS)×3
Zero Trust For Ai Agents — the vault's other agent-security framework (hub); AIMS composes existing…
- Autonomous Defense×3
Part V of Zero Trust For Ai Agents: securing the agents you deploy is only half the work — the…
- MCP and Computer Use×3
MCP security model: as the playbook prescribes wiring MCP into Salesforce, Gmail, Calendar for solo…
- Foundation → Enterprise → Advanced: Is the Agent Access-Control Jump a Cliff?×2
Not a cliff. The Enterprise tier is the pragmatic midpoint, and the open question that prompted…
- Agent Identity and Authentication×2
Identity and authentication form the foundation for every other security capability in Zero Trust…
- Agent Supply Chain Risk×2
Unlike static software supply chains, agentic ecosystems compose capabilities at runtime — loading…
- Agentic Prompt Injection×2
Prompt injection is the insertion of malicious instructions that cause an agent to follow attacker…
- AI-Accelerated Offense×2
Zero Trust For Ai Agents — the framework AI-accelerated offense motivates; it raised the Foundation…
- Anthropic×2
Zero Trust For Ai Agents — Anthropic's enterprise agent-security framework; positions Claude Code…
- Autonomous Intrusion×2
That is the datum. The threat model behind Ai Accelerated Offense and Zero Trust For Ai Agents was,…
- Claude Code×2
Zero Trust For Ai Agents cites Claude Code as the running example for nearly every control tier —…
- Impossible, Not Tedious (Design Test)×2
Zero Trust For Ai Agents — the framework that adopts this as a standing design-review question (hub)
- Least Agency×2
Zero Trust For Ai Agents — the framework that adopts least agency as its authorization principle…
- Open Questions Backlog×2
Zero Trust For Ai Agents ×3 (oldest 76d) — The framework treats every Claude Code "Pro-tip" as a…
- OWASP×2
The Open Worldwide Application Security Project — a long-standing nonprofit security community,…
- Standardize the Infrastructure, Not the Tools×2
The load-bearing phrase is the source's own summary of why this scales: "The infrastructure governs…
- Agent Data Injection (ADI)
Zero Trust For Ai Agents — ADI forges the very trusted anchors (origin metadata, tool history) Zero…
- Agentic Misalignment (AM)
Externally-induced analogue: Zero Trust For Ai Agents and Agentic Prompt Injection — Zero Trust…
- AI-to-AI Coercion
Zero Trust For Ai Agents — adjacent to but not on that page's agent–agent isolation boundary: the…
- Automated Failure Attribution
Zero Trust For Ai Agents — these numbers price a security agenda item, not just a reliability one.…
- Capability Gating Is Not Authorization
Zero Trust For Ai Agents — the concrete instantiation of the framework's Phase 5 "secure tool…
- Does 'Impossible, Not Tedious' Kill Defense-in-Depth? Layered Friction, Agent-Relativity, and the Frequency Paradox
So the test's verdict on defense-in-depth is a refinement, not a repeal: friction is demoted from…
- LLM-Driven Vulnerability Research
Ai Accelerated Offense — the threat-landscape generalization of these findings: vuln-to-exploit…
- MCP Tool Poisoning
Zero Trust For Ai Agents — a Phase-4/Phase-5 threat (input validation + secure tool access):…
- Agent Security
Zero Trust For Ai Agents (hub) — Anthropic's security framework for deploying autonomous agents:…
- Non-Malleable Memory Authority (TMA-NM)
Zero Trust For Ai Agents — the concrete construction behind the framework's Phase 7 "safeguard…
- Off-Host, Identity-Bound Authorization
Zero Trust For Ai Agents — a concrete instantiation of the framework's Phase 4 (defend against…
- Out-of-Band Prompt-Injection Defense
Zero Trust For Ai Agents — this is the academic-systems instantiation of the framework's Phase 4…
- Self-Propagating Prompt Injection (AI Worms)
Zero Trust For Ai Agents — "assume attacker-controlled content in context results in compromise at…
- Task-Specification Effects in Prompt Injection (AutoDojo)
Zero Trust For Ai Agents — a user–agent boundary entry in that page's five-boundary isolation map,…
- Write-Then-Trusted
Zero Trust For Ai Agents — Pillar's prescriptions are the framework's doctrine restated for the…
Related articles
- Least Agency
OWASP term extending least privilege to agents: constrain not just what an agent can access but what each tool can do,…
- Agentic Prompt Injection
Direct and indirect injection of malicious instructions into an agent; LLMs cannot reliably distinguish information fro…
- Blast Radius (Agentic)
The potential damage if an agent is compromised; the unit Zero Trust's 'assume breach' posture is built to contain via…
- Capability Gating Is Not Authorization
Agent frameworks ship capability gating (which tools are exposed, schema validity) but no fail-closed per-call authoriz…
- Agent Data Injection (ADI)
A new category of indirect prompt injection: malicious payloads disguised as *trusted data* (metadata like a comment's…
