Pillar AI Security Platform

Pillar Security is an AI security company co-founded by Dor Sarig and Ziv Karliner, with its contact address on Menachem Begin Road in Tel Aviv. It helps organizations discover, test, secure and govern AI agents, LLM applications, MCP servers and shadow AI, and published the SAIL framework as a practical guide to securing AI systems.

The platform has four parts. Discover and manage builds an inventory of agents, models, MCP servers and skills and applies AI posture management. Assess and red team maps exposure and runs agentic red teaming. Runtime guardrails detect and block prompt injection, jailbreaks, tool manipulation and leakage of card numbers, health records, passwords and API keys, while monitoring agent behaviour and logging every prompt, response and tool call to the SIEM. Governance maps policies to OWASP, MITRE ATLAS, NIST and the EU AI Act. It runs as cloud, self-hosted or hybrid, including in the customer's VPC.

• AI discovery and posture: Inventories agents, models, MCP servers and skills, including shadow AI.
• Agentic red teaming: Maps attack surface exposure and validates vulnerabilities.
• Runtime guardrails: Blocks prompt injection, jailbreaks and tool manipulation using intent analysis.
• Data privacy enforcement: Masks or blocks PII, PHI, credentials and API keys in prompts and responses.
• Agent behaviour monitoring: Flags drift from an agent's intended role and logs every tool call.
• Governance and compliance: Maps policies to OWASP, MITRE ATLAS, NIST and the EU AI Act.

An insurer in Vietnam can red-team its claims assistant before launch and then run guardrails that mask policyholder data. A software outsourcing firm can inventory the MCP servers and coding agents its engineers use and enforce permissions. A bank can stream AI agent audit logs into its SOC for incident response.

Guardrails are tuned with findings from red teaming of each specific application, so protection fits the real attack surface rather than generic rules. Sensitive data in prompts and responses can be masked, blocked or logged by policy. Full audit logs streamed to the SIEM support investigations and compliance evidence. VPC and self-hosted options keep regulated data inside the customer's environment.

AI coding assistants, MCP-connected agents and RAG applications have produced a steady stream of new vulnerabilities such as indirect prompt injection and sandbox escapes. Security teams need controls that understand agent behaviour, not just network traffic. Organizations in Vietnam and Southeast Asia building GenAI products face the same risks as regulators move toward AI and personal data rules.

Pillar Security operates through Pillar Security Inc. and Pillar Security Ltd, with an address in Tel Aviv-Yafo. It raised a 9 million USD seed round in April 2025, backed by Shield Capital and Golden Ventures. It holds SOC 2 Type II and ISO 27001 and was named a 2026 Gartner Cool Vendor in AI Software Security.