Pillar Security threat coverage: OWASP LLM Top 10 & MITRE ATLAS (2026)
ActivePillar Security is an AI-SPM / Runtime Protection vendor based in Miami, Florida, USA and Tel Aviv, Israel. Its flagship product is Pillar platform (AI Discovery & Posture, RedGraph red teaming, Runtime Guardrails, Agentic Endpoint with Runtime Guardian, Governance & Compliance). Against the OWASP LLM Top 10 (2026) we score it Covered on 5 of 10 risks, Partial on 2, with 6 MITRE ATLAS techniques mapped as of . It's independent.
- 5/10
- Risks covered
- 2
- Partial
- 6
- ATLAS techniques
- $9M
- Funding
- Active
- Status
At a glance
- Category
- AI-SPM / Runtime Protection
- Headquarters
- Miami, Florida, USA and Tel Aviv, Israel
- Founded
- 2023
- Employees
- 11-50
- Flagship product
- Pillar platform (AI Discovery & Posture, RedGraph red teaming, Runtime Guardrails, Agentic Endpoint with Runtime Guardian, Governance & Compliance)
- Deployment
- Managed cloud, self-hosted in your cloud account or data center, or hybrid; endpoint rollout via MDM; adaptive guardrails are managed cloud only as of 2026-10
- Best for
- Teams building their own AI apps and agents that want red-team findings fed into runtime guardrails, with a self-hosted option
- Funding to date
- $9M
- Last round
- Seed ($9M) · 2025-04
- Status
- Active
Facts as of .
What does Pillar Security do?
Runs the whole AI lifecycle on one platform: discovery, agentic red teaming, runtime guardrails, and pre-tool blocking on developer machines. It ships as Managed cloud, self-hosted in your cloud account or data center, or hybrid; endpoint rollout via MDM; adaptive guardrails are managed cloud only as of 2026-10, and it's built for teams building their own AI apps and agents that want red-team findings fed into runtime guardrails, with a self-hosted option. It's a weaker fit for buyers who need rate limits, cost controls or hallucination checks from their AI security vendor; none is documented.
Which OWASP LLM Top 10 risks does Pillar Security cover?
Pillar Security's strongest verdict is LLM01 Prompt Injection. The two tables below come from our editorial read of Pillar Security's public documentation as of (it's what the docs support, not a hands-on lab test), and this row also sits in the full matrix across all 27 vendors.
| OWASP LLM Risk | What it is | Pillar Security coverage | How it's addressed | Source |
|---|---|---|---|---|
| LLM01 Prompt Injection | User or hidden input overrides the model's rules or intended behavior. | Covered | Runtime Guardrails Injection Blocking stops malicious instructions in user or retrieved content. | https://www.pillar.security/platform/runtime-guardrails |
| LLM02 Sensitive Information Disclosure | The model leaks secrets, personal data, or proprietary content in its output. | Covered | Runtime Guardrails detect PII, PHI, secrets and credentials, then mask, block or log them under policy. | https://www.pillar.security/platform/runtime-guardrails |
| LLM03 Excessive Agency | An agent holds more permissions, tools, or autonomy than the task needs. | Covered | Agentic Endpoint audits permission rules and approval modes, and Tool Allow-Lists restrict which tools each MCP server may invoke. | https://www.pillar.security/platform/agentic-endpoint |
| LLM04 Supply Chain | Compromised models, datasets, plugins, or dependencies add risk before runtime. | Covered | AI Discovery & Posture traces dependency chains, validates model provenance, and stops unapproved models from deploying. | https://www.pillar.security/platform/ai-discovery-posture |
| LLM05 Data and Model Poisoning | Tampered training or fine-tuning data corrupts how the model behaves. | Unverified | Pillar names training data poisoning as a homegrown AI risk, but no named control over training or fine-tuning data was found. | https://www.pillar.security/solutions |
| LLM06 Unbounded Consumption | Uncontrolled requests drive cost, denial of service, or model extraction. | Unverified | No rate limit, token budget or cost control found on Pillar product pages as of 2026-10-30. | None found |
| LLM07 Misinformation | The model produces false or fabricated content that users act on. | Unverified | No hallucination or factual-accuracy control found on Pillar product pages as of 2026-10-30. | None found |
| LLM08 Hidden Context Exposure | Hidden instructions, retrieved policy text, or tool schemas get extracted, inferred, or reconstructed from the model. | Covered | Runtime Guardrails System Prompt Protection prevents extraction or manipulation of instructions. | https://www.pillar.security/platform/runtime-guardrails |
| LLM09 Vector and Embedding Weaknesses | Flaws in RAG stores let attackers poison, extract, or infer data. | Partial | Pillar scans RAG pipelines for vulnerabilities before production, but no runtime vector-store or embedding control is documented. | https://www.pillar.security/solutions |
| LLM10 Improper Output Handling | Downstream systems trust model output without checking it, enabling injection or code execution. | Partial | Automatic Masking redacts sensitive content before it reaches downstream systems, but no control for injection through output is named. | https://www.pillar.security/platform/runtime-guardrails |
| ATLAS Technique (ID) | Tactic | Pillar Security coverage | Notes | Source |
|---|---|---|---|---|
| LLM Prompt Injection (AML.T0051) | Execution | Covered | Runtime Guardrails detect and block prompt injection, including instructions in retrieved content. | https://www.pillar.security/platform/runtime-guardrails |
| LLM Jailbreak (AML.T0054) | Defense Evasion / Privilege Escalation | Covered | Runtime Guardrails detect and block jailbreaking attempts. | https://www.pillar.security/platform/runtime-guardrails |
| Exfiltration via AI User Interface (AML.T0136) | Exfiltration | Covered | Runtime Guardrails mask or block sensitive data in prompts and responses under policy. | https://www.pillar.security/platform/runtime-guardrails |
| Data from AI Services: System Instructions (AML.T0085.003) | Collection | Covered | System Prompt Protection prevents extraction of instructions. | https://www.pillar.security/platform/runtime-guardrails |
| AI Agent Tool Invocation (AML.T0053) | Execution / Privilege Escalation / Lateral Movement | Covered | Runtime Guardian evaluates every action in the agent pre-tool hook before it runs, on Claude Code, Codex CLI and Cursor. | https://www.pillar.security/platform/agentic-endpoint |
| Supply Chain Compromise (AML.T0010) | Initial Access | Covered | Pillar scans model provenance against an approved list and blocks deployment of an unverified model. | https://www.pillar.security/platform |
Which OWASP Agentic risks does Pillar Security cover?
4 of 10 Covered and 4 Partial against the OWASP Top 10 for Agentic Applications, scored from public documentation as of 2026-10-30.
| Agentic Risk | What it is | Pillar Security coverage | How it's addressed | Source |
|---|---|---|---|---|
| Agent Goal Hijack (ASI01) | Attackers redirect what an agent is trying to do, often through documents or other content it reads. | Covered | Injection Blocking stops malicious instructions in retrieved content, and Agent Behavior Analysis checks actions against the expected business role. | https://www.pillar.security/platform/runtime-guardrails |
| Tool Misuse and Exploitation (ASI02) | An agent uses a legitimate tool in a harmful way, or a manipulated tool call does damage. | Covered | Runtime Guardrails Tool Abuse Detection blocks unauthorized access to external systems, and tool invocations are tracked across agentic workflows. | https://www.pillar.security/platform/runtime-guardrails |
| Identity and Privilege Abuse (ASI03) | Agents inherit, reuse, or escalate credentials and permissions beyond what the task needs. | Partial | Pillar flags agents that inherit overly broad access and deviate from declared intent, which is detection and not privilege enforcement. | https://www.pillar.security/solutions |
| Agentic Supply Chain Vulnerabilities (ASI04) | Compromised MCP servers, plugins, skills, or agents enter the runtime. | Covered | Agentic Endpoint scores each MCP server per version against 23,000+ analyzed servers, and Tool Allow-Lists can block tools. | https://www.pillar.security/platform/agentic-endpoint |
| Unexpected Code Execution (ASI05) | An agent generates or runs code that gives an attacker execution. | Covered | Runtime Guardian blocks risky commands before they run on Claude Code, Codex CLI and Cursor, and Malicious Hook Patterns catches code-executing hooks and skills. | https://www.pillar.security/platform/agentic-endpoint |
| Memory and Context Poisoning (ASI06) | Attackers plant content in an agent's memory or context that shapes later decisions. | Unverified | SAIL lists agent memory and context poisoning as a risk, but no named product control for it was found. | https://www.pillar.security/sail/agent-memory-and-context-poisoning |
| Insecure Inter-Agent Communication (ASI07) | Messages between agents get spoofed, intercepted, or tampered with. | Partial | Pillar inspects inter-agent handoffs and intercepts poisoned instructions, but no spoofing or message-authentication control is documented. | https://www.pillar.security/solutions |
| Cascading Failures (ASI08) | One fault spreads across connected agents and systems. | Partial | Pillar catches instructions at the first hop in an orchestration chain, and RedGraph assesses blast radius, but non-malicious fault spread is not documented. | https://www.pillar.security/platform |
| Human-Agent Trust Exploitation (ASI09) | An agent's confident output talks a person into approving something harmful. | Unverified | Runtime Guardian Ask mode prompts a developer to confirm, but no control against manipulated human approval is named. | https://www.pillar.security/platform/agentic-endpoint |
| Rogue Agents (ASI10) | An agent drifts or is compromised and acts outside policy. | Partial | Agent Behavior Analysis detects drift from intended business purpose, and an agent shutdown action is not documented. | https://www.pillar.security/platform/runtime-guardrails |
Is Pillar Security independent, and how is it funded?
Pillar Security is an independent company as of . It has raised $9M to date, most recently a Seed ($9M) dated 2025-04. Lead investors: Shield Capital.
Pillar Security alternatives
The closest alternatives we track in AI-SPM / Runtime Protection are Apiiro, Cranium, HiddenLayer.
Frequently asked questions
What is Pillar Security used for?
Pillar Security is an AI-SPM / Runtime Protection vendor. Its flagship product is Pillar platform (AI Discovery & Posture, RedGraph red teaming, Runtime Guardrails, Agentic Endpoint with Runtime Guardian, Governance & Compliance). It ships as Managed cloud, self-hosted in your cloud account or data center, or hybrid; endpoint rollout via MDM; adaptive guardrails are managed cloud only as of 2026-10, and it's built for teams building their own AI apps and agents that want red-team findings fed into runtime guardrails, with a self-hosted option.
Is Pillar Security independent or acquired?
Pillar Security is an independent company as of 2026-10-30 and has not been acquired.
How many OWASP LLM Top 10 risks does Pillar Security cover?
We score Pillar Security as Covered on 5 of the ten OWASP LLM Top 10 (2026) risks, with 2 more Partial, as of 2026-10-30. The table on this page breaks down every risk, including the ones marked Not covered or Unverified.
How is Pillar Security deployed?
Pillar Security ships as Managed cloud, self-hosted in your cloud account or data center, or hybrid; endpoint rollout via MDM; adaptive guardrails are managed cloud only as of 2026-10.
What are the best alternatives to Pillar Security?
The closest AI-SPM / Runtime Protection alternatives llmthreat tracks are Apiiro, Cranium, HiddenLayer.