EU AI Act Demystified: Risk Tiers, August 2 Rules, and Future Requirements
EU AI Act Demystified: Risk Tiers, August 2 Rules, and Future Requirements
Format: 60-minute presentation and panel, followed by 15-20 minutes of interactive Q&A and networking
Target Audience: Early-stage to mid-stage AI founders, technical leads, product managers, and compliance advisors.
Event Agenda
Part 1: The Four EU AI Risk Tiers (15 Minutes)
Every builder must determine their system's regulatory category. Non-compliance carries financial penalties up to 35 million euros or 7% of global annual turnover.
Unacceptable Risk - Social scoring, subliminal behavioral manipulation, emotion recognition in workplaces/schools. Banned completely in the EU. Systems must not be developed or deployed.
High Risk - AI in recruitment, credit scoring, medical devices, educational admissions, critical infrastructure.Heavily regulated. Requires risk management, high-quality data auditing, logging, and human oversight.
Limited Risk - Chatbots, AI coding assistants, synthetic media generation, deepfakes.Active enforcement from August 2. Mandatory transparency disclosures and content watermarking.
Minimal Risk - Spam filters, AI video game NPC behaviour, basic search and recommendation engines. No mandatory regulatory burden under the AI Act; voluntary codes of conduct apply.
Part 2: Active Enforcement vs. Future Milestones (20 Minutes)
Currently Enforced Requirements
User Interaction Disclosures: Chatbots and interactive agents must explicitly notify users that they are interacting with an AI system.
AI literacy: EU AI Act Article 4 statutory literacy training.
Future Developments and Deadlines
AI Regulatory Sandboxes: National authorities must establish operational AI Sandboxes, giving startups a controlled framework to test high-risk models under legal guidance before commercial launch.
Full High-Risk System Compliance: High-risk standalone AI applications and safety components transition into mandatory compliance oversight over upcoming regulatory windows.
Part 3: Practical Takeaway: The AI Founder's Compliance Toolkit (20 Minutes)
Attendees will receive an actionable technical and operational framework:
Risk Classification Decision Tree: A simple logic flow to categorise your product's regulatory tier.
Part 4: Interactive Q&A and Community Discussion (15 - 20 Minutes)
Open discussion covering open-source model responsibilities, edge cases, overlapping GDPR rules, and technical implementation challenges.