OffSec AI-300 Advanced AI Red Teaming Training

• Introduction to Red Teaming AI Systems
  • Understand how artificial intelligence systems change the traditional attack surface. This module introduces the core concepts of AI cybersecurity, explains how adversaries target AI-enabled environments, and maps AI attacks to the red team lifecycle and modern cyber defense practices.
• Reconnaissance for AI Targets
  • Learn how to identify and map AI applications, machine learning components, and model infrastructure within a target environment. Students practice reconnaissance techniques used to discover AI assets, dependencies, and exposed services without alerting defenders.
• Attacking AI Agents
  • Explore offensive techniques for manipulating AI agents by abusing prompt instructions, memory systems, and tool integrations. This module demonstrates how attackers influence autonomous AI applications while maintaining stealth.
• Attacking Multi-Agent Systems and A2A Protocols
  • Analyze the architecture of multi-agent AI systems and learn how adversaries exploit trust relationships between agents. Students practice attacks such as message manipulation, agent impersonation, and workflow corruption.
• Exploiting RAG Pipelines
  • Examine how attackers compromise retrieval-augmented generation (RAG) systems by poisoning knowledge sources and manipulating retrieval layers to control model outputs.
• Attacking Embedding
  • Understand the role of embeddings in machine learning systems and perform attacks such as embedding inversion and information extraction to recover sensitive data from AI models.
• Attacking Model Context Protocol and Tool Surfaces
  • Explore how orchestration layers and AI tool integration frameworks can be abused to escalate privileges or execute unintended actions within AI systems.
• Supply Chain Attacks on AI/ML Systems
  • Learn how adversaries target the AI supply chain, including datasets, model weights, adapters, and dependencies. Students practice techniques used to introduce malicious artifacts into AI environments before deployment.
• AI Infrastructure and Deployment Exploits
  • Identify vulnerabilities in AI infrastructure, including cloud security platforms, model servers, and containerized machine learning workloads.
• Threat Modeling for AI-Enabled Targets
  • Develop strategies for identifying high-value AI assets, trust boundaries, and potential attack paths in complex AI environments while supporting risk management and improving threat detection capabilities.
• Assembling The Pieces – Capstone Red Team Engagement
  • Apply the techniques learned throughout the course during a full-spectrum red team engagement against a realistic enterprise AI environment, simulating how adversaries compromise production AI systems.

The training is ideal for:

• Penetration Testers
• Red Teamers
• Security Engineers
• Cybersecurity professionals looking to specialize in AI security
• AI Engineers and Developers wanting to learn practical techniques for identifying and mitigating AI cybersecurity risks.

AI-300 is an advanced AI cybersecurity course designed for learners with a strong foundation in cybersecurity. Students should have experience with:

• Penetration testing concepts
• Networking
• Linux and Windows systems
• Basic scripting
• A basic familiarity with AI systems or machine learning concepts, such as LLMs or generative AI applications, is helpful but not required.
• OSCP or equivalent hands-on experience is recommended.

• Identify and map attack surfaces across modern AI systems, including generative AI, LLM applications, and machine learning environments.
• Perform reconnaissance and threat detection and modeling for AI-enabled systems, identifying trust boundaries and high-value targets.
• Exploit vulnerabilities in AI agents and multi-agent systems, including prompt injection and memory manipulation attacks.
• Compromise RAG pipelines and vector databases through data poisoning and retrieval-layer manipulation.
• Conduct embedding attacks and extract sensitive information from AI models and machine learning systems.
• Exploit weaknesses in AI orchestration layers and tool integration frameworks used by modern AI applications.
• Identify and exploit vulnerabilities across the AI supply chain, including datasets, models, and adapters.
• Attack AI infrastructure and deployment environments, including model servers, cloud security platforms, and containerized workloads.
• Perform model extraction, adversarial machine learning attacks, and AI system manipulation techniques.
• Apply offensive methodology to assess AI cybersecurity risks and improve risk management strategies across AI environments.