- “AI Foundations Quick Start” – Building a Common Foundation of AI Knowledge
- Definition of AI and the difference between rule-based systems and machine learning (comparison table).
- Six everyday examples of AI (voice assistants, spam filters, video streaming recommendations, etc.).
- Key acronyms and terminology (ML, NLP, GPAI, HITL, etc.).
- Mini knowledge quiz (K1/K2).
- AI Fundamentals and the Scope of the AI Act
- Definition of an AI system according to the AI Act; categories and practical use cases.
- Risk categories: prohibited, high-risk, limited-risk, and minimal-risk systems.
- Roles and responsibilities (provider, deployer, importer, distributor).
- AI literacy obligations and the principle of proportionality.
- EU AI Act timeline: 2024 → 2027.
- Extraterritorial scope of the regulation (providers outside the EU).
- Opportunities and criticism of the regulation (innovation, SME compliance costs).
- Risk Identification and Human Oversight
- How to identify a high-risk AI system.
- Human oversight requirements and decision documentation.
- Cheat sheet: HITL / HOTL / HOOTL – definitions, examples, and intervention procedures.
- Impact assessments (DPIA/FRIA) – when and how to conduct them.
- Error and “hallucination” scenarios – escalation procedures.
- Responsibilities of Organizations and Users (Deployers)
- Minimum organizational requirements: policies, registers, and procedures.
- Documentation of training, authorizations, and access rights.
- Use case inventory register – sample template.
- Cooperation with AI vendors (SaaS, foundation models, APIs) – contracts and compliance considerations.
- Interaction with sector-specific regulations (MDR, PSD2, transportation, cybersecurity).
- Law and Ethics: GDPR + Transparency
- Data processing principles (lawful basis, data minimization).
- Privacy by Design / Privacy by Default and security requirements (Articles 25 and 32 GDPR).
- Non-discrimination, bias mitigation, and training data governance.
- “AI Transparency Playbook” (Article 52):
- Obligation to inform users when interacting with AI.
- Labeling AI-generated content.
- Disclosure of biometric data usage.
- Templates: chatbot notification banners and image watermarking.
- Relationship between the AI Act and GDPR – overlapping and distinct requirements.
- Prompt Engineering Fundamentals
- The LLM mental model: role, context, task, format, and evaluation criteria.
- Structure of a “good prompt” (RKTFF framework).
- Limitations, quality metrics, and token hygiene.
- Advanced Prompt Engineering Techniques and Business Scenarios
- Example-based learning, chain-of-thought prompting, and self-reflection techniques.
- Prompt chaining and orchestration (agent-based workflows).
- Prompt templates – versioning and corporate repositories.
- Best and worst practices: disclosure versus anonymization of business secrets and sensitive information.
- Validation, Quality Assessment, and Auditability
- Checklists and evaluation rubrics for AI model outputs.
- Traceability and prompt version control.
- Bias & Fairness Casebook – three scenarios (HR, lending, healthcare) with bias mitigation exercises.
- Compliance checklist (business, legal, IT, and security perspectives).
- Maintaining Compliance and Developing Competencies
- Mechanisms for continuous upskilling (regulatory, policy, and model updates).
- Training update plans – monitoring regulatory guidance and AI tool developments.
- Support channels: internal AI Office, knowledge base, and Q&A resources.
Sii Poland
SII UKRAINE
SII SWEDEN
