AI-SPM and AI-BOM vendors
You cannot secure an AI asset you have not found. AI security posture management (AI-SPM) discovers models, agents, AI services and shadow AI, checks them for misconfigurations and attack paths, and produces an AI bill of materials (AI-BOM). 34 of 56 vendors document it. Cloud and exposure platforms that document discovery:Orca Security, Qualys TotalAI, Tenable One AI Exposure, Trend Micro TrendAI Vision One and Wiz. Vendors that name an AI-BOM: Checkmarx LLM Scanner, Microsoft AI security, Upwind and Wiz.
34 vendors that document AI-SPM and AI-BOM
Checked 11 Nov 2026| Vendor | What it documents | Also covers |
|---|---|---|
| Arthur | Discovers agents on endpoints, cloud, and private data centers, including shadow agents. 1 | Guardrails, Agent security |
| Aurascape | Automatically detects and risk-scores AI applications, copilots, and embedded models. 1 | Guardrails, Agent security, Red teaming, Data protection |
| Cato Networks AI Security | Discovers and remediates lifecycle risks and scans internal AI model configurations. 1 | Guardrails |
| Checkmarx LLM Scanner | Produces deterministic AI-BOM evidence with model provenance, findings, and risk posture. 1 | Model scanning |
| Cisco AI Defense | Inventories AI models and connected data sources across distributed environments. 1 | Guardrails, Red teaming, Model scanning, Data protection |
| Credo AI | AI Registry catalogs AI systems, agents, vendors and models, with shadow AI discovery and risk classification. 1 | Agent security |
| CrowdStrike Falcon AI security | Falcon Guardian discovers known and shadow AI agents across endpoints, cloud, and SaaS. 1 | Agent security, Model scanning |
| Cyera AI Guardian | Discovers and catalogs AI models, agents, apps and pipelines, including shadow AI, and maps them to identities, data and access paths. 1 | Agent security, Red teaming, Data protection |
| Fiddler AI | Maintains a living inventory of connected agents and enforcement decisions. 1 | Guardrails, Data protection |
| Harmonic Security | Discovers how employees and agents use AI across desktop apps, CLI tools, IDEs, APIs, and MCP workflows. 1 | Agent security |
| Holistic AI | Discovers models, agents, and applications across cloud, code, and SaaS. 1 | Agent security, Red teaming |
| IBM Guardium Exposure Manager | Maps sensitive-data movement across AI agents, applications, databases, and endpoints. 1 | Data protection |
| Knostic | Discovers shadow AI and centrally governs agent policies and configurations. 1 | Agent security, Model scanning |
| Lasso Security | Inventories AI agents and applications, maps their components, and assesses misconfigurations. 1 | Guardrails, Red teaming |
| Microsoft AI security | Defender CSPM discovers cross-cloud AI BOMs, identifies misconfigurations and attack paths, and supports AI agents in Preview. 2 | Guardrails, Agent security, Data protection |
| Mindgard | Maps AI systems, models, agents, tools, and behaviors before attacks. 1 | Agent security, Red teaming |
| NeuralTrust | Maintains a real-time inventory of agents, connected tools, and execution workflows. 1 | Guardrails |
| Oligo Security | Discovers shadow AI components running in clusters. 1 | Agent security, Model scanning |
| Operant AI | Endpoint Protector discovers shadow AI, coding agents, and MCP clients on employee devices. 1 | Guardrails, Agent security |
| Orca Security | Inventories AI/ML assets and checks misconfigurations, exposed access keys, and shadow AI. 1 | Data protection |
| Palo Alto Networks Prisma AIRS | Provides AI posture management for training/inference data, agents, applications, and model access. 1 | Guardrails, Agent security, Red teaming, Model scanning |
| Pillar Security | Catalogs agents, models, prompts, tools, MCP servers, and skills through agentless integrations. 1 | Guardrails, Red teaming |
| PromptArmor | Finds which vendors use AI and scores AI-specific risk across 26 risk vectors, then monitors vendors for changes. 1 | Model scanning |
| Proofpoint AI Security | AI Access Security discovers AI tools active in the environment. 1 | Guardrails, Agent security, Data protection |
| Qualys TotalAI | Discovers sanctioned/shadow AI across clouds, SaaS, GPUs, endpoints, pipelines, models, agents, and MCP servers. 1 | Guardrails, Red teaming |
| SentinelOne Prompt Security | Inventories AI tools and code assistants, including shadow AI. 1 | Guardrails, Agent security, Red teaming |
| Straiker | Discover AI maps agents, MCP servers, and workflows with posture and misconfiguration monitoring. 1 | Guardrails, Agent security, Red teaming |
| Tenable One AI Exposure | Tracks employee/agent AI use and identifies unsafe platform settings and integrations. 1 | Guardrails |
| Trend Micro TrendAI Vision One | Inventories AI services, models, workloads, storage, entitlements, repositories, and images. 1 | Guardrails, Red teaming |
| Upwind | AI-SPM flags exposed model endpoints, unversioned models, permissive IAM roles and exposed keys; the AI-BOM inventories libraries, frameworks, model registries and cloud AI services. 1 | Agent security, Red teaming, Data protection |
| WitnessAI | Inventories AI applications, agents, and MCP servers. 1 | Guardrails, Agent security, Red teaming |
| Wiz | Builds an agentless AI-BOM and detects AI service misconfigurations and attack paths. 1 | Guardrails, Data protection |
| Zenity | Keeps a live inventory of agents across SaaS, custom and endpoint deployments and checks their configuration and permissions. 1 | Guardrails, Agent security |
| Zscaler AI Security | Discovers models, workflows, code repositories, RAG systems, and MCP servers. 1 | Guardrails, Red teaming |
Model and supply-chain scanning
11 vendors document scanning of model files, agent skills or the AI supply chain for tampering and malicious code. Protect AI's capabilities now contribute to Palo Alto Networks Prisma AIRS.
| Vendor | What it documents |
|---|---|
| Checkmarx LLM Scanner | Static scanning finds unsafe deserialization, command execution, trust-remote-code, and malicious pickle patterns without executing models. 1 |
| Cisco AI Defense | Includes AI supply-chain risk management for models and files. 1 |
| CrowdStrike Falcon AI security | Endpoint protection stops malicious packages and install scripts introduced by coding agents. 1 |
| DeepKeep | Model Scanning performs static and dynamic supply-chain checks. 1 |
| HiddenLayer | Scans third-party and proprietary models for hidden risks. 1 |
| Knostic | AgentMesh scans skills, MCP servers, and VS Code extensions for supply-chain threats. 1 |
| Noma Security | Finds agent supply-chain risks and misconfigurations. 1 |
| Oligo Security | Inspects model files and dependencies loaded into GPU/CPU memory for integrity risks. 1 |
| Palo Alto Networks Prisma AIRS | Scans third-party models for tampering, malicious scripts, and deserialization attacks. 1 |
| PromptArmor | Assesses new AI assets in the agent supply chain, including connector scope and data flows to AI subprocessors. 1 |
| Snyk Agent Scan | Scans agent supply chains for injection, tool poisoning, cross-origin escalation, and toxic flows. 1 |
What to ask before you shortlist
- Agentless or agent-based? Agentless discovery, as in Orca and Wiz, covers cloud estates fast. Endpoint discovery finds shadow AI on laptops.
- Which AI services does it recognise? Compare the integrations on each vendor page against the clouds and model providers you use.
- Does it produce an AI-BOM you can audit? Checkmarx documents AI-BOM evidence mapped to NIST AI RMF, the EU AI Act and ISO 42001.
- Does posture come with enforcement? Discovery that cannot act on what it finds adds a report, not protection. Pair it with runtime guardrails.