Understanding System Description: A Key Component of SOC 2 Compliance

One in five organizations has already suffered a breach traced back to shadow AI. Meanwhile, 63% of breached organizations either have no AI governance policy at all or are still drafting one. Below is a complete, copy-ready shadow AI policy template with twelve sections, plus guidance on adapting it for your company size, your industry, and the regulatory frameworks you answer to. The template assumes one hard truth up front: your employees are already using unapproved AI tools. A policy that pretends adoption hasn’t started yet fails on day one, so this one starts from the assumption that it has. What Is a Shadow AI Policy? A shadow AI policy is a formal document that defines how your organization discovers, evaluates, approves, and governs AI tools that employees adopt outside official IT channels. The term borrows from shadow IT, the older problem of unsanctioned software and hardware, but the AI version carries sharper risks: data pasted into a public model may be retained, used for training, or exposed in ways the organization can’t reverse. The policy does three jobs: it separates approved use from unapproved use, gives employees a fast and visible way to request new tools so the sanctioned route beats the workaround, and spells out what happens when someone crosses the line, including how the organization detects it and responds. Shadow AI Policy vs. General AI Acceptable Use Policy Many organizations already have an AI acceptable use policy (AUP) and assume it covers shadow AI. It usually doesn’t. An AUP tells employees how to behave inside approved tools. A shadow AI policy governs the tools themselves: which ones exist in your environment, which ones are allowed, and what happens with the rest. You need both. The AUP handles conduct; the shadow AI policy handles inventory and control. If you only have room for one document, fold the AUP’s data-handling rules into Section 6 of the template below. Let Axipro help you build a business continuity plan that’s practical, compliant, and audit-ready. Strengthen Your Business Continuity Strategy​ Schedule A Consultation The Shadow AI Policy Template (Download Link and Copy-Ready Sections) We’ve created a compliance safe template for Shadow AI Policy, use the link below to create a copy and customize for your company: Download The Shadow AI Policy Template → Copy the sections below into your policy management system and replace the bracketed placeholders. The language is plain on purpose. Legalese gets skimmed. Section 1: Purpose and Scope This policy governs the acquisition, approval, and use of artificial intelligence tools, features, and services at [Company]. It applies to all employees, contractors, interns, and third parties with access to [Company] systems or data. It covers standalone AI applications, AI features embedded in existing software, browser extensions, AI agents, APIs, and personal AI accounts used for work purposes, on both corporate and personal devices. The purpose of this policy is to enable productive AI use while protecting [Company] data, customers, and legal obligations. This policy does not prohibit AI. It prohibits ungoverned AI. That last sentence matters. Employees read the purpose statement first, and it decides whether they see the policy as an enabler or a blocker. Section 2: Definitions and Terminology Shadow AI: any AI tool, feature, agent, or service used for work purposes without formal approval under this policy. Approved AI Tool: an AI tool listed in the Approved AI Tools Registry (Section 4) and used under a [Company]-managed account. Personal AI Account: an account on any AI service registered to a personal email address or paid for personally. AI Feature: AI functionality embedded within otherwise approved software (e.g., an AI assistant added to a project management tool), which requires separate evaluation. Sensitive Data: data classified as [Confidential] or [Restricted] under [Company]‘s data classification policy, including the prohibited data classes in Section 6. Define “AI feature” explicitly. Vendors now ship AI additions into already-approved SaaS products every month, and without this definition, those features inherit approval they never earned. Section 3: Roles and Responsibilities The CISO (or designated security lead) owns this policy, maintains the Approved AI Tools Registry, and runs the approval workflow. Department heads ensure their teams know the policy and surface tool requests rather than suppressing them. Legal and Compliance review tools that touch regulated data or fall under the EU AI Act, GDPR, HIPAA, or client contractual restrictions. IT operates detection and monitoring controls (Section 9). Every employee is responsible for using only approved tools for work, reporting unapproved AI use they discover, and requesting new tools through the workflow in Section 7 rather than adopting them directly. Insider Note: In organizations under roughly 200 people, the “CISO” in this section is often the same overworked IT lead who manages laptops. Name a real person, not a title that doesn’t exist yet. A policy that assigns duties to a phantom role is unenforceable, and auditors notice. Section 4: Approved AI Tools Registry [Company] maintains a registry of approved AI tools at [location/URL]. For each tool, the registry records: tool name and vendor, approved use cases, prohibited use cases, permitted data classes, account type (enterprise/team/individual), data retention and training settings, risk tier (Section 5), approval date, and next review date. Only tools listed in the registry may be used for work. Tools not listed are unapproved by default. The registry is reviewed [quarterly]. Keep the registry somewhere employees actually look, such as your intranet homepage or IT help center, not buried in a GRC platform they can’t access. An invisible registry recreates the problem the policy exists to fix. Section 5: Risk Tier Classification (Low, Medium, High) Each tool in the registry is assigned a risk tier. Low: the tool processes only public or internal non-sensitive data, runs under an enterprise agreement with training opt-out, and produces output that a human reviews before use. Approval by IT Security alone. Medium: the tool processes internal business data or connects to [Company] systems via API or integration. Approval by IT Security plus the data owner. High: the

Legacy threat modeling frameworks such as STRIDE were designed for software that behaves the same way over and over again. Agentic AI does no such thing. It can rewrite its own plan mid-task, call external tools, negotiate with other agents, and produce a different output from identical input. MAESTRO exists because none of the legacy threat modeling frameworks were built to handle that. MAESTRO stands for Multi-Agent Environment, Security, Threat, Risk, and Outcome. It is a seven-layer threat modeling framework created specifically for agentic AI systems, and it has become the closest thing the industry has to a standard method for reasoning about agent security. Understanding MAESTRO in the Context of Agentic AI What MAESTRO Stands For Each word in the acronym carries meaning. Multi-Agent Environment signals that the framework models entire ecosystems of interacting agents, not a single model behind an API. Security, Threat, Risk covers the core discipline: identifying attack surfaces, cataloging threats, and assessing likelihood and impact. Outcome is the part most frameworks skip. MAESTRO asks what an attack actually produces in the real world, because an autonomous agent with tool access turns a compromised prompt into a compromised action. The Origin of MAESTRO (Cloud Security Alliance) The Cloud Security Alliance published MAESTRO in February 2025. Its creator is Ken Huang, Co-Chair of the CSA AI Safety Working Groups and CEO of DistributedApps.ai. The CSA has since applied the framework publicly to real systems, including OpenAI’s Responses API and Google’s A2A protocol, which gives practitioners worked examples rather than just theory. The framework is openly published, and the CSA maintains an official companion tool, the MAESTRO Threat Analyzer, on GitHub. SOC 2, ISO 27001 and HIPAA done for you. Fixed fee, 100% audit pass rate. Audit-ready in 6 weeks. Not 6 months. Schedule Free Assessment Why Traditional Frameworks Fall Short for Agentic AI STRIDE, PASTA, LINDDUN, and OCTAVE all share a founding assumption: the system under analysis follows predictable logic with clearly defined boundaries. You draw the data flow diagram, mark the trust boundaries, and enumerate threats against components that behave deterministically. Agentic AI breaks every part of that assumption. Unique Security Challenges of Autonomous Agents Agents introduce three properties that legacy models cannot express. Non-determinism means the same input can produce different behavior, so you cannot enumerate execution paths in advance. Autonomy means the agent makes decisions and takes actions without a human approving each step, which collapses the usual assumption that a person sits between intent and execution. And in multi-agent systems there is often no stable trust boundary: agents delegate to other agents, consume tool outputs from external servers via protocols like the Model Context Protocol (MCP), and update their own memory and goals at runtime. The Gap Between Legacy Frameworks and Agent-Based Systems The practical consequence is coverage gaps. STRIDE has no category for goal manipulation, where an attacker gradually steers what an agent is trying to achieve. PASTA assumes attacker objectives and data flows are fixed, which fails for systems that learn and adapt during operation. LINDDUN addresses privacy but says nothing about agent collusion or memory poisoning. A threat model built purely on these frameworks will pass review and still miss the attacks that matter most in an agentic deployment. How MAESTRO Addresses Agentic-Specific Risks MAESTRO does not discard the older frameworks. It extends them with a layered reference architecture, an AI-specific threat catalog for each layer, and, critically, explicit analysis of how threats propagate between layers. That cross-layer lens is the framework’s real contribution, because most serious agentic incidents are chains: poisoned data influences a model, the model misleads an agent, and the agent takes an unauthorized action three layers away from where the attack started. The Seven Layers of the MAESTRO Framework MAESTRO decomposes any agentic system into seven layers, each with its own threat landscape. Layer 1: Foundation Models The core LLMs or other models the agents reason with. Threats here include adversarial examples, model extraction, backdoored weights, and jailbreaks that bypass safety training. If the model is a third-party API, supply chain risk lives at this layer too. Layer 2: Data Operations Everything the agent ingests, stores, and retrieves: training data, RAG pipelines, vector databases, and agent memory. Data poisoning and memory tampering are the signature threats at this layer, and they are especially dangerous because a poisoned memory persists across sessions and keeps shaping future decisions long after the initial attack. Layer 3: Agent Frameworks The orchestration software that turns a model into an agent: LangChain, CrewAI, AutoGen, custom planners, and tool-calling logic. Threats include prompt injection through tool outputs, insecure tool definitions, and manipulation of the planning loop itself. Layer 4: Deployment Infrastructure The servers, containers, and cloud services the agents run on. The CSA’s threat catalog here reads like traditional cloud security with an agentic twist: compromised container images carrying malicious agent code, Kubernetes orchestration attacks, denial of service against agent runtimes, and tampering with Infrastructure-as-Code templates that provision agent resources. Layer 5: Evaluation and Observability The systems that monitor, evaluate, and debug agent behavior. This layer is often forgotten, and attackers know it. The CSA specifically flags poisoning observability data: manipulating the telemetry fed to monitoring systems so that incidents stay hidden from security teams while malicious activity continues. Layer 6: Security and Compliance MAESTRO treats this as a vertical layer that cuts across all others: identity and access management, guardrails, policy enforcement, and compliance controls. Threats include permission escalation, guardrail bypass, and compromise of the security agents themselves in architectures where AI enforces policy on other AI. Layer 7: Agent Ecosystem The environment where agents interact with users, other agents, and marketplaces. This is where the genuinely novel threats live: agent impersonation, misleading agent capability cards, tool squatting, and collusion between agents to achieve outcomes no single agent was authorized to pursue. Insider Note: In real assessments, Layers 5 and 6 expose the maturity gap fastest. Most teams’ shipping agents can describe their model and their orchestration framework in detail, then