This Auditing the Installation and Accreditation of GPAI Systems course guides the auditor with the planning and auditing of general-purpose AI after installation.

Course Overview

Drawing from the EU AI Act and the GPAI Code of Practice, the course explores a structured roadmap for developing AI deployment objectives—into practical, actionable insights. Whether you're auditing staff training, resource sizing, or post-implementation reviews, this course equips you with the understanding to ensure AI systems are not just operational but robust, secure, and aligned with regulatory mandates for safety, transparency, and fundamental rights protection.

Through real-world examples from leading models like OpenAI's GPT, Google's Gemini, Anthropic's Claude, Meta's Llama, and xAI's Grok, you'll explore how to mitigate risks such as biases, cyber offenses, loss of control, and societal harms (e.g., misinformation or CBRN threats). This course bridges technical deployment with legal and ethical responsibilities, offering a shortcut to mastering the intricate mechanisms that foster a responsible AI ecosystem.

This course will equip participants with a sharper perspective on organizational pitfalls, emphasizing proactive strategies to ensure integrity, compliance, and reliability in AI initiatives.

Learning Objectives

By the end of this course, participants will be able to:

• Identify and describe nine core categories of AI risk drivers, including change management, transitions, service agreements, operational performance, resilience, incident handling, configuration, problem resolution, and data governance.
• Understand how risks like model drift, bias amplification, unauthorized changes, data poisoning, and resource misallocation interconnect and cascade across AI systems.
• Recognize the role of human factors, legal compliance (e.g., GDPR, EU AI Act), ethical considerations, and environmental impacts in mitigating AI vulnerabilities.
• Apply key takeaways to evaluate and address real-world challenges in developing, deploying, and maintaining AI, fostering a culture of continuous vigilance and adaptation.

Key Topics Covered:

The course systematically unpacks each phase of the AI Five Process, integrating EU AI Act requirements (e.g., Articles 4, 9, 10, 14, 15, 27, 43, and 72) and GPAI Code of Practice commitments for safety, security, and transparency:

• Staff Training and AI Literacy: Embedding ethical considerations, risk identification, and human oversight; auditing proficiency through assessments and policies.
• Application Software Performance Sizing: Forecasting computational resources (e.g., FLOPs, energy consumption) with sustainability in mind; evaluating benchmarks like MLPerf for robustness.
• Implementation Planning: Developing roadmaps with risk management, trigger points for interventions, and progress metrics (e.g., KPIs, milestone rates).
• System and Data Conversion: Ensuring compatibility, integrity, and accuracy in transitions; addressing multimodal data, bias mitigation, and lawful content access per Article 10.
• Testing Strategies and Plans: Designing comprehensive verification, including red teaming for vulnerabilities; covering conformity assessments under Article 43.
• Testing Changes and Modifications: Independent reviews of updates with risk reassessments (Article 9) and cybersecurity mitigations.
• Parallel, Pilot, and Operational Testing: Simulating production environments with predefined criteria; validating robustness and usage tracking.
• Final Acceptance and Security Accreditation: Formal evaluations by users and IT; accrediting security levels with metrics like CVSS scores.
• Promotion to Production: Authorized migrations in segregated environments; ensuring acceptable risks per Code of Practice Measure 4.2.
• Post-Implementation Evaluations: Assessing user requirements, fundamental rights impacts (Article 27), and overall success through surveys and logs.
• Management's Post-Implementation Review: Conducting root cause analyses for incidents; applying lessons for continuous improvement and post-market monitoring (Article 72).

Target Audience: This course is ideal for AI developers, project managers, compliance officers, IT professionals, auditors, and anyone involved in AI adoption or oversight. Familiarity with basic AI concepts is beneficial.