OSAI — Advanced AI Red Teaming (AI-300)

Section 01

01. At a Glance

About the course

OSAI is OffSec's AI Red Teamer certification — a 24-hour practical exam against an AI-enabled enterprise where you attack LLM agents, multi-agent protocols, RAG pipelines, embeddings, MCP tool surfaces and the infrastructure underneath. Macksofy's bootcamp pairs the official AI-300 course with the LearnToExploit Model Attack & Red Teaming lab track, so every attack class is practised before exam day.

Section 02

02. Who Is This Course For

• OSCP / OSEP holders moving into AI red teaming
• Penetration testers asked to assess LLM and agent features
• AI engineers who want to understand adversarial threats to their own systems
• Security engineers defending AI infrastructure

Prerequisites

• OSCP or equivalent hands-on penetration testing experience (OffSec's recommendation)
• Comfort with Linux, Windows, networking and basic scripting
• No prior AI / ML background required — helpful, not mandatory

Section 03

03. What You Will Be Able To Do

• Fingerprint an AI target: model, tools, retrieval, guardrails and agent topology
• Compromise AI agents through direct and indirect prompt injection into tool calls
• Attack multi-agent systems and the Agent-to-Agent (A2A) protocol
• Exploit RAG pipelines, vector stores and embedding inversion
• Abuse Model Context Protocol (MCP) servers and tool surfaces
• Run AI/ML supply-chain attacks and pivot from AI infrastructure into cloud and Kubernetes
• Pass the 24-hour proctored OSAI exam and deliver the report

Section 04

04. Curriculum — 13 Modules

Module structure and topic coverage authored by Macksofy Technologies based on the publicly-published vendor syllabus, current as of the issue date of this brochure. Vendor reserves the right to revise content; Macksofy keeps cohort material aligned to the latest release.

1. Module 01

   Module 01 · Introduction to Red Teaming AI Systems

   3 topics
   • How LLM applications are assembled
   • Why instructions and data share one token stream
   • Scoping and rules of engagement for AI targets
2. Module 02

   Module 02 · Reconnaissance for AI Targets

   3 topics
   • Model and provider fingerprinting
   • Capability, tool and retrieval enumeration
   • Guardrail and filter mapping
3. Module 03

   Module 03 · Attacking AI Agents

   3 topics
   • Direct and indirect prompt injection into agent loops
   • Tool-call hijacking and excessive agency
   • Agent memory attacks
4. Module 04

   Module 04 · Attacking Multi-Agent Systems & the A2A Protocol

   3 topics
   • Agent discovery and registration abuse
   • Rogue-agent routing
   • Cross-agent trust and task injection
5. Module 05

   Module 05 · Exploiting RAG Pipelines

   3 topics
   • Retrieval access-control bypass
   • Stored injection through indexed documents
   • Knowledge-base poisoning
6. Module 06

   Module 06 · Attacking Embeddings

   3 topics
   • Embedding inversion (vec2text)
   • Vector-store exfiltration
   • Similarity-search manipulation
7. Module 07

   Module 07 · Attacking Model Context Protocol & Tool Surfaces

   3 topics
   • MCP server enumeration
   • Tool poisoning and description injection
   • Over-permissioned tool chains
8. Module 08

   Module 08 · Supply Chain Attacks on AI/ML Systems

   3 topics
   • Malicious model artefacts (pickle)
   • Tokenizer and model tampering
   • Module hijack in ML pipelines
9. Module 09

   Module 09 · AI Infrastructure & Deployment Exploits

   3 topics
   • Cloud IAM abuse from AI services
   • Kubernetes RBAC and container escape
   • GPU and inference-server exposure
10. Module 10

    Module 10 · Threat Modeling for AI-Enabled Targets

    3 topics
    • Trust-boundary inventory
    • MITRE ATLAS and OWASP LLM Top 10 mapping
    • Prioritising attack paths
11. Module 11

    Module 11 · Assembling the Pieces — Capstone Red Team Engagement

    3 topics
    • End-to-end chain: recon → app exploit → platform → persistence
    • Evidence for non-deterministic findings
    • Exam-style reporting
12. Module 12

    Macksofy bootcamp · LearnToExploit Model Attack & Red Teaming lab track

    3 topics
    • 12 modules and 9 browser labs mapped to the AI-300 modules
    • Jailbreak, output-handling and system-prompt extraction classes OSAI does not teach
    • Full-engagement reporting module
13. Module 13

    Macksofy bootcamp · 24-hour mock exam + report review

    3 topics
    • One timed mock engagement in an AI-enabled lab
    • Mentor review of the written report
    • Time-allocation playbook for the proctored exam

Section 05

05. Tools You Will Operate

Section 06

06. Career Outcomes

Section 07

07. Placement Support

Macksofy's placement desk works directly with hiring partners across India and the UAE. Resume coaching, mock interviews and direct intros included.

• 1:1 resume + LinkedIn rewrite with our hiring desk
• Mock interviews with active practitioners
• Direct intros to BFSI, fintech and Big-4 partners
• UAE placement support (Dubai, Abu Dhabi)

Section 08

08. Why Macksofy

• Vendor-true delivery — Macksofy is a hands-on cybersecurity training provider delivering practitioner-led bootcamps with exam-prep support.
• Practitioner-led delivery — every Macksofy instructor is a working OSCP / OSWE / OSEP / CISA-certified consultant on real client engagements during the week.
• Mentor support until you pass — extended access to mentor office hours and exam-day prep at no additional cost.
• Placement desk — Macksofy works with hiring partners across India and the UAE; your post-course resume, portfolio review and mock interviews are included.
• Indian classroom + online cohorts — onsite delivery in Mumbai BKC and Hyderabad HITEC City; live virtual cohorts pan-India with recordings.

Section 09

09. How to Enrol

1. Submit the enquiry form at macksofy.com/contact or call +91 99308 24239.
2. A Macksofy advisor will respond within 4 business hours with the next batch dates, payment terms and invoice.
3. Confirm enrolment via NEFT / RTGS / corporate card. EMI options available for select courses.
4. Receive welcome kit, lab credentials and the cohort calendar within 24 hours of confirmation.