Lasso Security protects the way enterprises use, build and ship AI. It is for security teams who need to see which AI tools employees are using, which agents are running with access to company tools and data, and which customer-facing AI applications could be manipulated, and then to put policy around all three.
The problem is that AI adoption has outrun the controls around it. Data leaves in prompts, agents act with more access than they need, and public assistants can be talked into misbehaving. Lasso gives security teams one place to see and govern all of it.
At A Glance
| Vendor | Lasso Security |
|---|---|
| Category | AI Model Security, AI Safety and Guardrails |
| Identities covered | AI agents and their tools, MCP server connections, employee AI use |
| Deployment | Gateway, API or SDK, browser extension, proxy integration |
| Agentic AI | Agent discovery, intent-based runtime protection, MCP Gateway |
| Customers | US Department of Homeland Security, BMW, eToro, Fiverr, Kaltura |
| Founded | 2023 |
| Market classification | AI Model Security; also AI Safety & Guardrails |
What Lasso Security Does
Enterprise AI brings three kinds of exposure. Employees paste data into AI tools nobody approved. Agents act on their own with credentials and tool access that may be far broader than they need. Customer-facing chatbots and assistants face adversarial prompts from outside. Lasso handles all three with the same approach: understand how each agent or application is put together before it goes live, attack it as an adversary would, and then judge each action in production by its intent, so a legitimate tool call can be told apart from one that has been hijacked. Lasso reports more than 1 million AI security threats mitigated in its first year and a half.
Products
AI Usage Control
Security teams worried about shadow AI use this to see which AI tools staff are using, and to apply data loss prevention policies before sensitive data leaves in a prompt.
AI Agents Security
As agents multiply, few organisations know how many they have or what each can touch. This product governs agents from build to runtime:
- Discovers and inventories every agent and its tools, with a risk score for each
- Flags overprivileged or high-risk agents
- Sets which tools an agent may use and which resources it may access
- Protects agents at runtime at the intent layer and keeps an audit log of their autonomous decisions
AI Application Protection
Teams shipping chatbots, back-end AI services and assistants to customers get inline guardrails that block malicious input and unsafe output, backed by automated red teaming that simulates attacks against LLMs and AI applications before release, so weaknesses surface in testing rather than in production.
AI Security Posture Management
AI-SPM builds a graph of each agent: its models, tools, APIs, databases, MCP servers, sub-agents, prompts and execution paths. From that graph it finds misconfigurations, weak guardrails, exposed dependencies, attack paths, excessive permissions and unsafe tool use. An AI bill of materials is assembled from code repositories, cloud environments and low-code platforms.
Intent Security And LEAP
Intent Security is the analysis engine behind real-time detection. LEAP is a “transformer-free” guardrail that runs on ordinary server processors without GPUs, with a reported verdict time under five milliseconds. A second engine, RAPID, handles decisions that need more judgment.
MCP Gateway
The open-source MCP Gateway sits as a proxy and orchestrator for MCP interactions. It filters requests and responses through guardrails and adds usage analytics and a dashboard across MCP servers.
How It Works
Lasso is delivered through a gateway, an API or an SDK, and can be started with a single line of code.
| Component | What it does |
|---|---|
| Runtime enforcement | Policies run inline at the proxy, API or AI gateway layer. |
| Discovery | Agents on commercial platforms are found through platform integrations, and homegrown applications by connecting through CI. |
| Employee controls | A browser extension or proxy integration covers staff use of AI tools. |
| Policy templates | Pre-built templates follow OWASP, MITRE and NIST guidance. |
| Data integrations | Findings flow into the customer’s existing security tools, so AI alerts are triaged alongside everything else. |
Identities And Systems Covered
| AI agents | Agents and the tools they use, including overprivileged agents |
|---|---|
| MCP servers | Connections between agents and MCP servers |
| Employees | Staff use of AI tools, including shadow AI |
| Agent platforms | Vertex AI, Microsoft Copilot, AWS Bedrock, Microsoft Azure and Salesforce Agentforce |
| Build environments | Code repositories, CI/CD pipelines, cloud environments and low-code platforms |
Integrations
| Area | Integrations |
|---|---|
| Security stack | Palo Alto Networks, CrowdStrike Falcon, Cloudflare |
| Identity | Okta (SSO) |
Company Background
| Founded | 2023, by four founders. |
|---|---|
| Headquarters | Tel Aviv, Israel, with US operations. |
| Leadership | Elad Schulman (CEO), Ophir Dror (CPO), Lior Ziv (CTO) and Yuval Abadi (COO). |
| Funding | $30 million announced on 2 September 2026, led by ClearSky Advisors with Entrée Capital, iAngels, Singtel Innov8, Mindset and Swish Data, for a reported total of more than $37 million. CyberArk, Mindful Ventures and Selah Ventures are also backers. |
| Recognition | Included in Gartner Cool Vendors in AI Security. |
Note: This overview is compiled from Lasso Security’s publicly available information as of 3 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.