[Guides]

Playbooks for AI governance.

Practical, vendor-neutral guides to help you discover, secure, and govern AI across your company.

[More about AI]

The Agentic AI Security Framework

Our open model for securing autonomous AI across models, MCP, the browser, and coding agents.

Featured white paper

A control model for autonomous AI

Agents do not just answer questions, they read data, call tools, and act on systems. The Agentic AI Security Framework defines the surfaces, threats, and controls every organization needs to govern them, mapped to ISO 42001, the EU AI Act, and SOC 2.

Read the framework
Cerbera

White paper

Agentic AI Security Framework

v1.0 · Cerbera Research

[Library]

Guides and playbooks.

What is AIDR? AI Detection and Response

Why the AIDR category emerged, what it requires at runtime, and how to adopt it without a platform migration.

Sébastien DucSébastien Duc

AI security categories, compared

A neutral map of AI-SPM, AI firewalls, AI DLP, AI governance, and AIDR, plus five questions to ask any vendor.

Arnaud DrizardArnaud Drizard

The shadow AI discovery playbook

A step-by-step method to inventory every AI tool, agent, and MCP server in your org in under a week.

Arnaud DrizardArnaud Drizard

Securing AI coding agents

How to let engineers use Cursor, Claude Code, and Copilot without exfiltrating source or secrets.

Robin CosteRobin Coste

Vetting MCP servers before they connect

A risk-scoring rubric for evaluating MCP servers and detecting rogue ones in flight.

Robin CosteRobin Coste

Building an AI acceptable-use policy

A template and rollout plan that keeps teams fast while satisfying security and legal.

Robin CosteRobin Coste

Getting ready for ISO 42001

What the AI management system standard expects, and the controls you can deploy today.

Robin CosteRobin Coste

DLP for the AI era

Why prompt-level data loss prevention matters and how to deploy it across browser, IDE, and CLI.

Sébastien DucSébastien Duc

The MCP supply chain: a new class of risk

How autonomous agents inherit the permissions of every MCP server they touch, and a model for governing them.

Sébastien DucSébastien Duc

Prompt-level DLP for the AI era

Why traditional data loss prevention misses AI traffic, and how transparent proxying closes the gap.

Robin CosteRobin Coste

Measuring shadow AI

A behavioral methodology for discovering unsanctioned AI tools across an organization.

Arnaud DrizardArnaud Drizard

Obscurity was your access control

Stale tokens, undocumented endpoints, and unreviewed internal tools were safe because nobody was looking. Agents look at everything and act on what they find.

Sébastien DucSébastien Duc