Workshops: EU AI Act & Compliance
Understanding EU AI Act Risk Classes & High-Risk AI
In this training you learn to classify AI applications confidently under the EU AI Act's risk categories. Using many real-world cases, you practise classification, from prohibited practices and high-risk systems to applications with transparency duties or minimal risk, and you understand the consequences of each outcome.
Who is this workshop for?
- AI leads and governance teams
- Compliance, data protection and risk management
- Product management and buyers of AI solutions
- Business units with potentially sensitive use cases such as HR, lending or education
- Consultants and internal auditors
What you will take away
- You know the four risk levels and the special rules for general-purpose AI models.
- You recognise prohibited practices such as certain forms of social scoring, manipulation or emotion recognition in the workplace.
- You map applications to the high-risk areas in Annex III and the product rules in Annex I.
- You check whether the exemption for systems without significant risk may apply, and document the result.
- You know the obligations a high-risk classification triggers for providers and deployers.
Agenda
Module 1
The risk pyramid
- Unacceptable, high, limited and minimal risk
- Where general-purpose AI models fit
Module 2
Prohibited practices
- Manipulation, exploitation of vulnerabilities, social scoring
- Emotion recognition in workplaces and education
- Biometric categorisation and remote identification
Module 3
High-risk AI in detail
- Annex I: AI as a safety component of regulated products
- Annex III: biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration, justice
- Exemptions for preparatory or narrow tasks, and their limits where profiling is involved
Module 4
Transparency and minimal risk
- Chatbots, deepfakes, synthetic content
- What applies to most office applications
Module 5
Case workshop
- Exercise: classify and justify 15 practical cases
- Consequences of classification and documenting the result
Why classification decides everything
Whether an AI application triggers hardly any duties or extensive requirements for risk management, documentation and oversight depends on its classification. Mistakes are costly in both directions: classifying too low carries fine and liability risk, classifying too high ties up resources unnecessarily.
Classification depends less on the technology than on the intended purpose. The same language model, such as ChatGPT, Claude, Gemini or Mistral, can mean minimal risk in marketing and a high-risk system in candidate selection.
Learning from cases
The core of the training is a case workshop with examples from different industries. You are welcome to bring your own applications, which we classify together in anonymised form.
- AI-assisted pre-screening of job applications
- Credit scoring for consumers
- Customer service chatbot
- Automated grading of exams
- Microsoft 365 Copilot for internal documents
- Predictive maintenance in an industrial plant
Caution with borderline cases
The European Commission is refining classification through guidelines and examples, and interpretation continues to evolve. We present the current state and a traceable method. For borderline cases with significant consequences, we recommend consulting your legal advisers.
Frequently asked questions
Is ChatGPT a high-risk system?
Not as such. The underlying models are subject to their own rules for general-purpose AI models. Whether a specific application is high-risk depends on its intended purpose, for example if it is used for recruitment decisions.
Who decides on the classification?
Primarily the provider of the system, who must document its assessment when relying on an exemption. Deployers should still check the classification of the systems they use, because it determines their own obligations.
How does this differ from the fundamentals training?
The fundamentals training gives the overall picture. This training focuses solely on classification and practises it intensively with cases.
Workshop
Get in touch
Shall we classify your own AI applications in the case workshop?
Tell us a little about your team – we will get back to you within 24 hours with a proposal for content, format and dates.
Prefer to talk?
More workshops
- Workshops: EU AI Act & ComplianceEU AI Act for HR: Workshop on AI in Recruiting & Performance
- Workshops: EU AI Act & ComplianceEU AI Act and GDPR Together: Combined Compliance Workshop
- Workshops: EU AI Act & ComplianceWrite Your Company AI Policy: Half-Day Workshop
- AI Development ServicesFixed Scope vs. Time and Materials for AI Projects
