Hands-on AI security training

Red team AI systems before attackers do.

Practice on live AI systems in your own private lab. Concise lessons grounded in real attacks, then exploit them yourself in minutes.

Browse the labs

2 labs free
No card
Nothing to install

Attack labs on live LLMs

Guided lessons

10/10

OWASP LLM Top 10 covered

Curriculum levels

How it works

Theory, then the real thing.

No slideware. Read the attack, then run it against a live model in a sandbox that's yours — with an AI mentor on hand when you're stuck.

Learn the theory

Concise lessons on each attack class — prompt injection, RAG poisoning, agentic abuse — grounded in real 2026 breaches.

Attack real LLMs

Launch a private lab and exploit an actual model. Capture the flag, with your AI mentor watching the whole session.

Earn your certificate

Clear the tracks to earn the verifiable KageX AI Red Teamer (KART) certificate.

The attack surface

Every layer is a target. Every target is a lab.

An AI product is a model wired to prompts, documents, tools, an app and the network behind it. Each wire is a way in. Pick one to see the path an attack takes.

What you get

Everything between reading about it and doing it.

20 hosted labs

Attack labs

Live LLMs, RAG stacks and agents in a private sandbox. Nothing to install.

Browse the labs → 9 levels · 54 lessons

Curriculum

Attack-first lessons across the OWASP LLM Top 10, Agentic Top 10, MITRE ATLAS, MCP vulnerabilities and live CVE tracking.

Open the Academy → Add-on

AI Mentor

A Claude tutor that has read the lesson and can see your lab.

See plans → Verifiable

KART certificate

Clear the tracks, earn a certificate with a public verification link.

How it is earned →

Mentor Lesson 2.3 · Indirect prompt injection PII Leakage lab open

youWhy does the model follow the instruction hidden in the retrieved page?
mentor reads Lesson 2.3 sees PII Leakage · 2/3 flags Because retrieval puts the page into the same context as your prompt, so the model can't tell your instruction from the page's. You hit exactly this in the PII Leakage lab when the summary leaked the salary table. Wrap the page in a delimiter and ask again; watch what still gets through.

Illustrative session · AI Mentor · scoped to your lesson and lab

AI Mentor

A tutor that already knows where you are.

A Claude tutor that has read every lesson and can see your labs, so you never explain yourself to a blank chat window.

Knows where you are: No; Lesson and lab, always
Sees your progress: No; Every solved and skipped lab
Offensive security: Refuses without setup; Defensive framing pinned in

Same Claude model. Different context.

Questions

What people ask before they start.

What is FreakLabs?

FreakLabs is KageX's hands-on AI security training platform. You attack live LLMs, RAG systems and AI agents in private cloud sandboxes, capture the flag, and work through a 9-level, 54-lesson curriculum. Two labs and the first two levels are free.

What is AI red teaming?

AI red teaming is deliberately attacking an AI system — prompt injection, jailbreaks, RAG poisoning, agent hijacking — to find what it will do that it should not, before an attacker does. It is the offensive half of AI security.

Do I need my own model or API key?

No. Every lab runs a real model in a sandbox spun up for you and torn down afterwards. There is nothing to install and no keys to supply.

Is FreakLabs free?

Two attack labs and the first two curriculum levels are free, with no card. The Individual plan is $29.99 a month and unlocks every lab, every lesson and the KART certificate.

Which frameworks does the curriculum cover?

The OWASP LLM Top 10, the Agentic Top 10, MITRE ATLAS and MCP vulnerabilities, plus live CVE tracking — across 9 levels and 54 lessons, each paired with a lab.

Who is FreakLabs for?

Security engineers, penetration testers, AppSec and ML engineers, and developers shipping LLM features: anyone who has to defend an AI system and has never attacked one.

Start free

Break your first AI today.

Two labs and the first two course levels are free — no card required. Upgrade when you're ready for the full attack catalog.

Browse the curriculum

Why FreakLabs

AI shipped faster than security knows how to defend it.

Every company is putting LLMs and agents in production. The security teams behind them have never red-teamed one. FreakLabs trains operators to break AI systems on purpose, so the people defending them know what to look for. You can't secure what you can't attack.

01 · The problem

AI is everywhere. Almost nobody knows how to secure it.

$280K

Top-end AI security salaries, but most cybersecurity professionals aren't qualified to fill these roles

35%

Surge in AI red-teaming demand projected by 2028 - and almost no training pipeline to meet it

600+

Contributors to the OWASP LLM Top 10 - massive community interest, but no structured learning path

02 · Why it matters

The training market hasn't caught up to AI

Established platforms (HackTheBox, OffSec, PortSwigger) ship excellent training for traditional web and network security, and have started adding AI labs. None ship the full track FreakLabs does: structured curriculum, sandboxed live LLMs, hours-fresh intel, and an AI mentor that reads the lesson with you.

Existing AI training

$8,000 three-day instructor-led courses (not repeatable)
Video-first: watch someone else red team AI
One prompt-injection module buried inside a 500-lab catalog
One-shot $1,499 certifications with no ongoing practice
No live mentor; you debug your jailbreaks alone

FreakLabs

20 attack labs, each in its own isolated cloud sandbox
Hands-on from day one. You attack, not watch.
9-level curriculum covering the full 2026 AI threat landscape
AI mentor with cross-lesson memory + hours-fresh CVE feed
Verifiable certificate that proves you can actually red-team AI

03 · Why now

The attack surface is exploding. Defenders are scarce.

The shift Every company is now an AI company

LLMs, copilots, and AI agents are being deployed into production across every industry, most without any adversarial testing.

The gap Security teams aren't trained for this

Traditional pen testing doesn't cover prompt injection, RAG poisoning, or agent hijacking. Existing training is passive, expensive, or scattered.

The opportunity AI red teamers are the most in-demand role in security

Companies are hiring faster than the talent pool can grow. Those who build this skill now will define the field for the next decade.

Your move Start attacking AI in a real sandbox today

FreakLabs gives you isolated cloud environments with live LLMs. Concise lessons paired with hands-on attack labs, ready in minutes.

Pricing

Break real AI systems.
For less than a coffee a day.

20 live targets: chatbots, RAG pipelines and agents, each in its own sandbox. Get in, take the flag, and leave with a certificate that proves it.

Start free, no card

Free forever: 2 levels and 2 labs
Cancel anytime
Prices in USD

What you walk away with

Skills you have used, not just read about

Real attacks on real models

Prompt injection, jailbreaks, RAG poisoning, agent hijacking. Every lab is a live model in a sandbox built for you.

A certificate that checks out

Clear 70% of the labs to earn KART. Every certificate has a public link anyone can use to verify it.

A mentor that reads along

The AI Mentor add-on walks through each lesson with you, checks your understanding and remembers what you have covered.

Questions

Before you pay

What do I get for free?

Two attack labs and the first two curriculum levels, each lab in its own private sandbox, plus 50 AI Mentor messages a month. No card, no time limit.

Can I cancel?

Yes, anytime, from your profile. You keep access until the end of the period you paid for, and your completed labs stay on your record.

What is the AI Mentor add-on?

A tutor that walks through each lesson with you, checks your understanding and remembers what you have covered. It is billed on top of any plan.

How does Team work?

One plan, four seats. Everyone gets every lab and lesson, and you see the team's progress and leaderboard in one place.

Do I need my own model or API key?

No. Every lab runs a real model in a sandbox we spin up for you and tear down afterwards. Nothing to install.

We need SSO, invoicing or more seats.

That is Enterprise. Email [email protected] and we will scope seats, SSO and custom labs with you.

Your first two targets are free.

No card. Break them, then decide.

Start free

Sessions

Active sessions

Resume a running lab or stop it before starting the next one.

Red team AI systems before attackers do.

Theory, then the real thing.

Learn the theory

Attack real LLMs

Earn your certificate

Every layer is a target. Every target is a lab.

Everything between reading about it and doing it.

Attack labs

Curriculum

AI Mentor

KART certificate

What's hitting AI systems this week.

A tutor that already knows where you are.

What people ask before they start.

What is FreakLabs?

What is AI red teaming?

Do I need my own model or API key?

Is FreakLabs free?

Which frameworks does the curriculum cover?

Who is FreakLabs for?

Break your first AI today.

Welcome back to FreakLabs.

Your progress

Recent sessions

AI shipped faster than security knows how to defend it.

AI is everywhere. Almost nobody knows how to secure it.

The training market hasn't caught up to AI

The attack surface is exploding. Defenders are scarce.

Preparing your lab

Sign in to FreakLabs

Break real AI systems.For less than a coffee a day.

Skills you have used, not just read about

Real attacks on real models

A certificate that checks out

A mentor that reads along

Before you pay

What do I get for free?

Can I cancel?

What is the AI Mentor add-on?

How does Team work?

Do I need my own model or API key?

We need SSO, invoicing or more seats.

Your first two targets are free.

Active sessions

Public profile

Account & security

KART certificate

Completed labs

Top performers

Nothing at .

Your first target is ready.

Break real AI systems.
For less than a coffee a day.