Asenion Test Console

What to test

Options

How many

Start

AI System to Test

What do you want to test?

Choose the type of test that matches your goal.

🛡️ OWASP Top 10 for LLM LLM vulnerabilities (prompt injection, data leaks) 🔐 OWASP Top 10 for Agentic AI Agentic applications (ASI01–ASI10) ⚖️ Bias & fairness Test for unfair or discriminatory behavior 👥 Multiturn Fairness Conversations with different user types (race, age, etc.) 📋 ISO/IEC 42001 AI management system compliance checks 🏛️ EU AI Act European AI regulation compliance checks

Options

Optional: narrow the focus.

Focus area (optional)

Vary by

Conversation length

Category (optional)

Runs all OWASP Top 10 for Agentic Applications (ASI01–ASI10). No options to narrow.

How many tests? (Goal: find vulnerabilities)

We run enough tests per category so you can be confident we've found vulnerabilities (95% confidence, ≤1% failure-rate bound). The exact number is shown on the next step.

Smoke test 1 per category · quick check only Light pass 5 per category · … Balanced 10 per category · … Find vulnerabilities (recommended) Confidence-based sample size · adaptive runs

Ready to start

1. Test type

Security and compliance tests, or fairness conversations with varied demographics.

2. System context

System purpose

Generic Auto-detect Manual

Run generic tests without detecting system purpose. Query the target to auto-detect purpose for domain-specific tests. Enter your system description below.

Describe your system Describe what your system does and its intended use case. This helps generate domain-specific tests.

3. Configuration

Testing Framework

OWASP Top 10 for LLM

LLM vulnerabilities

OWASP Top 10 for Agentic AI

Agentic applications (ASI01–ASI10)

ISO/IEC 42001

AI Management System

EU AI Act

European compliance

Category (Optional)

Tests Per Category

Adaptive Mode Re-runs successful attacks multiple rounds to verify consistency (no cache)

Rounds run until convergence. Steady State 10 (default), up to 100 for higher confidence. Docs

Steady State Rounds Consecutive passes for convergence (1–100)

Sample size calculator

Choose a goal to get the right recommended test count. Sample size doc

How to choose: Use Estimate pass rate when you want to report a precise pass rate (e.g. 85% ± 2%). Use Confidence we've found all vulnerabilities when you want to run tests, see zero failures, and then claim the true failure rate is below a bound (e.g. ≤1%). They answer different questions; neither is wrong.

Goal

Confidence level

Margin of error

Expected pass rate

Confidence level

Max acceptable failure rate

Run multi-turn conversations with the target using a simulated user (persona). One sensitive demographic feature is varied across tests to compare target behavior.

Vary feature Only this attribute changes across tests.

Number of turns Conversation length (user + assistant exchanges). 1–20.

Tests per variant (optional)

Number of Personas Total tests to run (1–1000). Updated when Tests per variant or Vary feature changes. Tests are allocated evenly across variants (round-robin).

Designated persona (optional) Base persona is generated from this description, then the selected vary feature is varied across tests. Leave empty to use a random job-candidate persona.

Parity metrics to run

Select one or more metrics to compute and report. Unselected metrics are omitted from results.

Assistant turns User turns Questions Sentiment Encouragement Follow-up depth Turn length ratio Pass rate (outcome)

Use memory-backed simulation Grounds the simulated user's persona and responses in real human experiences for more realistic conversations.

4. Target

AI System to Test

Loading...

Configured in providers/target.yaml

5. Run options

Run timeout (hours) Max run duration before timeout (1–168 hours). Default 3. Applies to both Red Team and Multiturn Fairness runs.

New Test Run

What do you want to test?

Options

How many tests? (Goal: find vulnerabilities)

Ready to start

1. Test type

2. System context

3. Configuration

4. Target

5. Run options

Configuration

Target Providers

Run History

Current Run

Run details

Result files

Delete test run?

Rename test run

Run Results

Upload to Test Repository

Validate AI Responses