EU AI Act — what does it require?

hard

Answer

$Risk-based classification: (1) Unacceptable risk (social scoring, subliminal manipulation): banned.$
$(2) High-risk (hiring, credit, law enforcement, education): requires risk management, data governance, transparency, human oversight, robustness testing, registration.$
$(3) Limited risk (chatbots, deepfakes): disclosure requirements.$
$(4) Minimal risk: free.$
$GPAI (Llama / GPT scale): transparency, copyright compliance, safety evaluations.$
$Penalties: up to 7% global revenue.$
$Timeline: phased 2024-2026.$
$Impacts every ML system used in EU.$

Check yourself — multiple choice

Random
Risk-based: unacceptable banned / high (hiring/credit/LE): risk mgmt+governance+transparency+oversight+robust+register / limited: disclose / minimal: free; GPAI: transparency+copyright+safety evals; penalties 7% global rev; phased 2024-2026
No rules
Not real

EU AI Act: risk-based; high-risk registered + audited; GPAI transparent.

#safety#llmops

215 interview questions in this topic.