Safety Layer for AI powered enterprise
We are building foundational tooling for AI Safety
We build the tooling that security teams, platform engineers, and AI operators need to understand, monitor, and protect the AI surface in their organizations.
Three tools. One mission.
Built from our research into how AI systems propagate through organizations — often invisibly. Each product targets a distinct blind spot in your AI security posture.
Radar
Connects to your network stack and discovers every AI tool in use across your org — not just the ones IT approved.
Surface
An endpoint agent that scans the assets network tools can't see — AI skills, MCP servers, local agents, and browser extensions.
| Asset | Type | Risk |
|---|---|---|
| github-copilot | VS Code ext. | low |
| claude-dev | MCP client | low |
| gpt-pilot-local | Local agent | med |
| unnamed-mcp-7 | MCP server | high |
| sider-ai | Browser ext. | high |
AIRiskDB
Threat intelligence built for the AI era. The largest structured database of AI tools and MCP servers, with risk profiles, permission scopes, and known vulnerability data.
The AI attack surface is growing faster than visibility into it.
AI tools have proliferated across every organization — on devices, in browsers, embedded into workflows. Most security teams have no idea what's running on their own networks.
Traditional security tooling was built for a world where software is installed deliberately and visibly. AI tools spread differently.
We believe safety and capability must be built together, not traded off. SuperAlign builds the foundational tooling so organizations don't have to choose.
You cannot protect what you cannot see. Every product we build starts with giving operators accurate, real-time visibility into their AI surface.
Traditional CVE frameworks weren't built for AI risk. We're building the threat intelligence infrastructure the field needs.
Every product we ship is grounded in original research. We find the problems before building the solutions.
Our tools are designed to work in real security operations centers, not just in research papers.