SANS Institute · AI security · Advanced
SEC536: Adversarial AI - Penetration Testing AI Systems (training course)
A hands-on course on attacking LLM applications, AI infrastructure, agents and MCP servers. It is a course, not a certification; the page does not name a linked exam.
- Who it is for
- Penetration testers and red teamers, application security engineers and architects, data scientists and ML platform engineers, blue teamers and technical managers.
- Prerequisites
- Comfort with HTTP APIs and general web and application security. AI/ML background and Python are helpful, not required.
- Format
- 3 days instructor-led, or 18 hours self-paced OnDemand with 4 months of access. 15 full hands-on labs.
- Cost
- US$5,250 (OnDemand and most live events); other regions priced separately on the page (e.g. S$6,815, EUR 4,935). Excludes local taxes.
- Renewal
- Not stated on the provider page. Check the provider.
What it covers
- LLM input processing, AI reconnaissance, direct and indirect prompt injection
- Memory and RAG poisoning, multimodal evasion
- Jailbreaking and guardrail evasion
- Exposed AI infrastructure, API authorization flaws, model weight extraction
- Alignment failure modes and agentic systems
- SSRF and sandbox escape, MCP server attacks
The GIAC GAIPT page links to this course, but GAIPT is presale only and is not listed here.
Official SANS Institute page →
Practise the skills
Practice only: PlayCISO does not award or accredit this credential.
Compare with
Sources (read 2026-11-08)
Details change. Confirm cost, eligibility and renewal with the provider before you pay.