Skip to content
BLACK BAG SECURITY
PEOPLE / PERSPECTIVE / PROGRESS
A security workstation with a browser validation workflow and evidence checks on a laptop screen.

AI Research to Practice

AI Browser
Security Test Suite

Test browser-based AI controls. Build safer systems.

A practical methodology and open test suite for validating AI solutions that operate through web browsers. Turn research into repeatable security testing, measurable outcomes, and stronger, safer deployments.

View on GitHub Read the Research

BROWSER
AI SECURITY
REAL THREATS
REAL TESTS
SAFER OUTCOMES

Core Principles

A structured approach to browser-based AI security.

CONTROLLED INPUTS

Test with realistic, adversarial browser content in a safe, repeatable environment.

MULTI-VIEW EVIDENCE

Capture DOM, rendered content, network, model context, and decisions.

POLICY OUTSIDE THE MODEL

Validate outputs with deterministic security policy, not model trust.

REPLAYABLE PROOF

Produce evidence packages that can be independently reviewed and reproduced.

The Methodology

From browser content to
measurable security outcomes.

Execute a complete security pipeline - from hostile browser content, through model interaction and validation, to enforceable security decisions with repeatable evidence.

Explore the Methodology
  1. Browser
    Artifact
  2. Evidence
    Capture
  3. Model
    Context
  4. Output
    Validation
  5. Deterministic
    Policy
  6. Analyst
    Review

EVIDENCEVALIDATIONPOLICYCONFIDENCE

Practical Lab Track

Hands-on. Repeatable. Realistic.

Step through 13 practical labs, from environment setup to a full target-backed evaluation, using real-world browser attack techniques and defensible validation methods.

Explore the Lab Track

13 LABS

Progressive, hands-on exercises (00 - 12)

REAL-WORLD

Attack techniques and AI browser scenarios

ACTIONABLE

Evidence and findings you can apply

Built For

Security teams. Product builders.
A safer AI ecosystem.

A practical resource for red teams, product security, SOC analysts, AI security researchers, and organizations deploying AI solutions that interact with web content.

Who It's For

HIGHER
STANDARDS
A MORE
RESILIENT WORLD

Research Origin

Grounded in research.
Built for practice.

The AI Browser Security Test Suite implements the Browser-Safe AI Systems research series, turning methodology into a practical, open test suite.

Read the Research Series
Part 24Red Team Testing Methodology Part 25Practical Lab Track Part 26Evidence and Validation

OPEN METHODS
PRACTICAL TOOLS
SAFER AI
TOGETHER