This Auditing General-Purpose AI Systems course provides auditors with a framework for navigating the complexities of GPAI and GPAIS, ensuring business objectives are fulfilled alongside safe, transparent and compliant utilization.

Course Description: Auditing General-Purpose AI

Overview

This advanced course provides essential training on auditing general-purpose AI (GPAI) models—versatile AI systems like large language models capable of performing a wide range of tasks without specific fine-tuning, posing unique challenges such as systemic risks, biases, and transparency issues. Effective auditing ensures compliance, mitigates harms, and promotes trustworthy deployment in applications from content generation to decision support. Aligned with the EU AI Act's requirements for GPAI models—applicable since August 2, 2025, including technical documentation, copyright compliance, training data summaries, and enhanced obligations for systemic-risk models like model evaluations, adversarial testing, incident reporting, and cybersecurity measures—ISO/IEC 42001:2023 for AI management systems emphasizing governance, transparency, and continuous improvement, ISO/IEC 42006:2025 for consistent auditing and certification of AI systems, and best practices including the GPAI Code of Practice, risk assessments across data-model-deployment, automated evidence collection, and continuous assurance, participants will learn to conduct audits that support innovation while safeguarding societal interests in fields like enterprise software and public services.

Learning Objectives

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

• Audit GPAI models for EU AI Act compliance, including transparency requirements (e.g., technical documentation, training data summaries, copyright adherence) and systemic-risk measures (e.g., evaluations, testing, reporting) effective since August 2025.
• Implement ISO/IEC 42001 and ISO/IEC 42006 standards to assess AI management systems for GPAI, ensuring consistent auditing, certification, impartiality, and competence in governance and risk treatment.
• Apply best practices for GPAI auditing, such as the GPAI Code of Practice, comprehensive assessments targeting data-model-deployment, automated tools for evidence and controls testing, and continuous assurance mechanisms.
• Develop audit strategies that integrate regulatory conformance, ethical considerations, and innovation support for GPAI, addressing challenges like open-source models and systemic risks.

Target Audience

This course is designed for auditors, compliance professionals, AI ethicists, risk managers, and developers working with GPAI models. It is particularly relevant for providers and deployers in EU-regulated markets or those seeking global certification for versatile AI systems.

Course Overview

AI technologies using General Purpose AI (GPAI) and GPAI with Systemic Risk (GPAISR) present deployers with unique challenges, complex risks, and compliance demands. This course provides auditors with a structured framework for auditing the deployment of GPAI systems.

Learning Objectives

This course will assist:

• Define and validate business, compliance, and data quality requirements for GPAI systems.
• Evaluate deployment strategies, and assess technical/economic feasibility.
• Consider the data architecture, risk analysis, and security measures.
• Determine traceability, user-centric design, procurement governance, and provider support.
• Manage systemic risks, copyright compliance, and data dependencies for trustworthy AI.

Course Structure

The course is divided into modules, each focusing on key auditing and deployment phases with real-world examples (e.g., fraud detection, healthcare AI).

Foundational Planning and Requirements

• Defining business, compliance, and data quality needs (e.g., EU AI Act, GDPR).
• Risks, controls, and audits for foundational requirements.

Evaluating Alternatives and Deployment Strategy

• Assessing proprietary vs. open-source solutions and provider compliance.
• Formulating strategies for acquisition, customization, and stakeholder alignment.

Provider Engagement and Technical Compatibility

• Specifying third-party requirements via RFPs and due diligence.
• Assessing integration, scalability, and human oversight.

Economic Feasibility and Data Architecture

• Conducting cost-benefit analyses, including non-monetary impacts.
• Establishing enterprise data models, integrity, and security.

Risk Analysis, Security, and Traceability

• Identifying risks, mitigations, and documentation.
• Implementing cost-effective security and audit trails.

User Interaction, Software Selection, and Procurement

• Ensuring ergonomics, transparency, and user instructions.
• Selecting system software and governing procurement.

Acquisition, Support, and Customization

• Acquiring products with compliance verification.
• Ensuring ongoing provider support and contracting custom programming.

Infrastructure Acceptance and Specialized Risks

• Accepting facilities and technology with performance/security tests.
• Evaluating systemic risks, copyright policies, and data dependencies.

Duration and Format

• Duration: 20 minutes.
• Format: Interactive sessions with lectures, group discussions, quizzes, and a capstone project on auditing a GPAI model.
• Assessment: Certificate awarded upon 70% quiz score and project completion.

Expected Outcomes

Participants will be ready to audit general-purpose AI models in compliance with the EU AI Act's requirements, ISO standards, and best practices, enabling organizations to manage systemic risks, ensure transparency, and drive responsible AI innovation.