For pentesters

AI is in your scope now

The scoping call mentions a chatbot, an agent with tool access, a RAG (retrieval-augmented generation) pipeline. You can test the web and cloud layer around it already — this closes the part that is new, on live targets rather than slideware. It's a repeatable approach to testing AI applications: identify the vulnerability, validate its impact, and document the finding.

Live models, not simulationsGraded on captured evidenceBuy it yourself — no employer needed

Buying for a practice or a team?

The gap

Your instincts transfer. Your techniques don't.

Payload lists alone aren't enough

Injection here is a persuasion problem, not a syntax one. The same request phrased two ways gets refused once and obeyed once — so you learn the application's behaviour and trust boundaries, not just a list to paste.

The boundary is not where you expect

A prompt is not a parser. Trust breaks between retrieval and context, or between an agent and its tools — often in a different file.

Impact is hard to argue

“The model said something rude” is not a finding. Turning behaviour into demonstrated business impact is the skill clients pay for.

The report is different

Non-determinism means a client will re-run your payload and get a different answer. Evidence has to be captured, not described.

Coverage

Add LLM features, agents, and MCP servers to the systems you can test

Work every category in the OWASP LLM Top 10, then MCP (Model Context Protocol) tool servers and agentic systems, against deliberately vulnerable applications running live. Use the web, API and cloud judgement you already have to test prompt injection, retrieval boundaries, tool access, and agent behaviour.

genaisecuritylab.com/learn-ai-security

Your scope, this year

Web appYou already test this

APIYou already test this

CloudYou already test this

LLM featuresNew in scope

Agents & toolsNew in scope

MCP serversNew in scope

Method

Five steps. Same order. Every engagement.

Same five steps whether the client ships a chatbot or an agent, so your second AI engagement does not start from scratch.

genaisecuritylab.com/methodology

AI pentest methodology

Scope

Map the AI surface: models, tools, retrieval, agents

Map

Trace where untrusted text reaches context

Test

Work the OWASP LLM Top 10 against each entry point

Verify

Re-run the payload against the proposed fix

Report

Evidence, impact, remediation the client can action

Then prove it with GSCE

GSCE — the GenAI Security Certified Expert ($399) — is hands-on, proctored through your webcam, and graded on the exploits you capture, with a few lines from you on how you found each and an examiner reviewing the sitting before the credential is issued. Weighted 75% GenAI, 20% web, 5% cloud — the exam's design, focused on where AI-security work concentrates. GSCP — the GenAI Security Certified Professional ($199) — is the intermediate hands-on exam to it.

Paying for it yourself

You don't need an employer to sign off

Individual

$19/mo · $228 billed yearly

or $29/mo billed monthly

Every lab and the full curriculum. The GSCE sitting is $399 on its own, whenever you are ready — or GSCP, the intermediate exam, at $199.

GSCE certification

$399one-time · 3 months of labs included

Launch price until 27 November, then $499.

The GSCE exam, proctored and examiner-reviewed, plus three months of lab access to prepare — included. One free retake, then $99 per attempt. There's no attendance version of this credential — you pass it or you don't. Still building AI-specific technique? GSCP, the intermediate exam, is $199 and the recommended step before GSCE.

Break a live model before your next scoping call

The free tier is 25 labs against a live model, no card required.