cybermap.lol
Defenses · 453 companies · 14 classes

The AI-security landscape, by problem class.

Every incident on cybermap.lol is tagged with the classes of tooling that work on its kind of problem. This page is the directory behind those tags. Nothing here is a verdict on any incident. Paid placements are labelled Sponsored; HeimWall, which runs this site, is labelled Operator.

Observability & Governance

297 companies

Inventory and behavioral visibility over every AI agent and assistant in use, with audit trails and policy. The class of tooling that answers "what did the agents do, and was it allowed".

Runtime & Guardrails

231 companies

Inline inspection at inference time: prompt injection, jailbreaks, data exfiltration and unauthorized tool calls stopped before they reach the model or a downstream system.

AI Data Governance

76 companies

Knowing which sensitive data (secrets, PII, regulated records) flows into models, assistants and agents, and controlling it.

AI Identity

116 companies

Scoped, verifiable, revocable identities and least-privilege access for agents and non-human workloads; stolen or over-broad credentials are the common thread in most breaches.

MCP & LLM Gateways

107 companies

A control point between agents and the tools and models they call: policy, logging and rate control for MCP servers and LLM traffic.

AI CSPM

73 companies

Posture management for AI infrastructure in the cloud: exposed endpoints, misconfigured services, unmanaged models and keys.

AI Endpoint Security

42 companies

Protecting the machines where people and agents work: malware, ransomware payloads and AI-assisted tooling on the endpoint.

AI Network Security

25 companies

Traffic-level defense: DDoS, command-and-control, lateral movement and AI-driven network attacks.

AI Code Security

23 companies

Securing what AI-assisted developers ship: poisoned packages, insecure generated code, supply-chain integrity.

Sandboxing

25 companies

Isolated execution for agents and untrusted code so a compromised tool cannot reach the rest of the system.

Model Security

56 companies

Protecting models themselves: weights, training data, tampering and theft.

AI Red Teaming

94 companies

Adversarial testing of models, agents and assistants before attackers do it.

AI Browser Security

33 companies

Guarding AI use inside the browser: extensions, web agents and data pasted into web assistants.

AI SSPM

15 companies

Posture and access control across the SaaS and AI apps an organization has adopted, sanctioned or not.

Vendor data: AI Security Startups Map by Prompt Security (CC BY 4.0), synced daily. startups.prompt.security · last sync 2026-10-11. Category descriptions are ours. To be listed or to sponsor a class, email founders@heimwall.ai.