Advanced AI Red Teaming (AI-300) – OSAI+ Certification
Training including examination and certification
.svg)
Machine Learning Engineer
- Fundamentals and Methods of AI Red Teaming
- Reconnaissance and Identification of AI Vulnerabilities
- Attacks on Generative AI and Large Language Models (LLMs)
- AI Agents: Prompt-, Memory-, and Tool-Based Attacks
- Multi-Agent Systems and Agent-to-Agent (A2A) Protocols
- Retrieval-Augmented Generation (RAG): Manipulation and Poisoning of Knowledge Sources and Retrieval Processes
- Attacks on Embeddings and the Extraction of Sensitive Information
- Model Context Protocol (MCP) and Attacks on Tool Integrations
- Supply-chain attacks on AI/ML systems, models, data, and dependencies
- Vulnerabilities in AI infrastructures, cloud AI platforms, model servers, and containerized ML workloads
- Threat Modeling for AI-Based Systems
- Combination of various attack techniques in a realistic AI red team engagement
- Hands-on exercises and labs on modern AI attack scenarios
- Preparation for the OffSec AI Red Teamer (OSAI+) Certification Exam
Consists of the following modules:
- Advanced AI Red Teaming (AI-300) – OSAI+ Certification
- Brush-Up: Advanced AI Red Teaming
Requirements:
Since the AI-300 is an advanced-level training course, you must have a solid technical background and practical experience:
- Strong knowledge of penetration testing and offensive security
- Confident Use of Linux, Networks, and the Command Line
- Strong knowledge of Python and experience reading and modifying code
- Basic understanding of web application security, APIs, and modern application architectures
- Understanding of fundamental AI and machine learning concepts, particularly large language models (LLMs) and generative AI
- Experience working with AI applications and LLM-based systems is a plus
To successfully earn certification, it's important that you thoroughly engage with complex technical problems and practice independently in the labs.
Recommendation: Practical experience at the PEN-200/OSCP level or comparable knowledge will make it easier for you to get started.
Technical Requirements
For the training and the subsequent exam, you'll need a high-performance computer of your own on which a virtual machine can run locally. The device must meet the following requirements:
- at least 8 GB of RAM
- At least 80 GB of free disk space
- the latest version of VMware Workstation or VMware Fusion
- Administrator privileges or sufficient permissions to install the virtual machine and configure system settings
- Support for x86/x64 VMware VMs
ARM-based devices, such as Apple Silicon Macs with M1, M2, M3, or newer M processors, are not supported. VMware Player and VMware Server are also not suitable for this training due to functional limitations.
- You identify and analyze attack surfaces and vulnerabilities in modern AI systems.
- You assess the security risks of LLMs, RAG applications, and AI agents.
- You plan and execute structured red-team attacks on AI-based applications.
- You exploit AI-specific vulnerabilities and misconfigurations in realistic attack scenarios.
- You manipulate AI agents, models, data sources, and connected tools to investigate potential attack vectors.
- You combine various attack techniques across applications, models, the supply chain, and infrastructure.
- You will document and assess identified vulnerabilities and attack vectors as part of a professional AI red team engagement.
The course combines five days of intensive, instructor-led training with a subsequent individual learning and practice phase. During the training days, the AI-300 curriculum is covered in a structured manner, and the theory is consistently integrated with demonstrations, hands-on exercises, and realistic AI red-teaming scenarios.
After the training, you’ll reinforce what you’ve learned on your own with your Learn One subscription. For 365 days, you’ll have access to OffSec’s learning materials and labs to solidify your understanding of attack techniques, solve complex scenarios on your own, and further develop your practical skills in a targeted manner.
Upon request, participants have the option to attend a half-day refresher session after the training concludes to reinforce their knowledge in preparation for the exam. The refresher session takes place on a separate date to be scheduled after the training. There is no additional cost for participation.
The Learn One subscription includes:
- 365 days of access to AI-300 and the accompanying hands-on labs
- AI-300 course materials, including the option to download the PDF and video versions during the subscription period
- Two attempts at the OSAI+ certification exam
- Access to Proving Grounds (PG Play & Practice) for additional hands-on training
- Access to the learning paths and Fundamentals content included in Learn One
- Access to PEN-210 (Wireless Attacks), including one OSWP exam attempt
- Access to PEN-103 / Kali Linux, including one KLCP exam attempt
- Baseline Assessments and Assessments of the 100-Level Content
The exam fees are included in the seminar price.
You have a total of two attempts at the corresponding certification exam.
With the AI-300, you can prepare for the OffSec AI Red Team Certification (OSAI+). The exam is hands-on and tests your ability to identify and exploit vulnerabilities in modern AI systems and to document your findings professionally.
- Exam duration: 23 hours and 45 minutes
- Documentation: Up to 24 hours thereafter to submit the audit report
- Format: practical, supervised hands-on exam
- Test Environment: Realistic AI Red Teaming Scenarios
- Focus Areas: Attacks on AI applications and systems, as well as the combination of various AI red-teaming techniques
- Audit Report: The procedure, identified weaknesses, and results achieved must be documented in a manner that is easy to follow
- Practice Exams: Two practice exams are included in your Learn One subscription
Additional Requirements for the Online Exam
The exam will be administered online and under supervision. You will also need:
- a built-in or external webcam; smartphones and tablets are not permitted as webcams
- A stable Internet connection with a download speed of at least 20 Mbit/s and an upload speed of at least 10 Mbit/s
- a supported browser, such as Google Chrome, Firefox, Chromium, Brave, or Microsoft Edge
- a private exam room
- a valid government-issued photo ID with information clearly visible in English
The webcam must remain on throughout the entire exam and be positioned so that your face, upper body, and immediate surroundings are visible. Before the exam, your identity, your workspace, the programs running on your computer, and your technical setup will be verified.
Please review the current OffSec requirements well in advance of the training and exam: OffSec Proctored Exam Requirements
Certification
Upon successful completion of the exam, you will receive the OffSec AI Red Team Certification (OSAI+) and the corresponding digital certificate from OffSec. No additional application or proof of specific work experience is required.
Recertification
OSAI+ certification is valid for three years. To maintain “+” status, OffSec offers various recertification options, including earning the required Continuing Professional Education (CPE) credits or passing a qualifying OffSec exam. The current provisions of the OffSec Continuing Professional Education Program apply.
The AI-300 is ideal for you if you already have a solid foundation in cybersecurity, penetration testing, or AI/ML and want to specialize in offensive testing of modern AI systems. The course is specifically designed for:
Penetration Testers & Red Teamers: You want to expand your existing offensive security skills to include LLMs, AI agents, RAG systems, and other modern AI technologies, and explore new attack vectors through hands-on testing.
AI Security Specialists & Security Engineers: You are responsible for the security of AI-based systems and want to understand attack techniques from an attacker’s perspective, identify vulnerabilities, and conduct thorough assessments of security risks.
AI/ML Engineers & Developers: You develop or maintain AI applications and want to understand how models, data, agents, tools, and AI infrastructures can be attacked and manipulated.
Security Consultants & Experienced Cybersecurity Professionals: You want to expand your skill set to include AI red teaming and demonstrate your practical skills with the OffSec AI Red Teamer Certification (OSAI+).
.svg)