AI regulation AI compliance & EU AI Act — meet European regulation
The AI Act is in force — are you compliant? Inventory of your AI systems, risk classification, legal obligations and compliance plan. Up to €35M in fines for the most serious breaches.
What you get
AI Act compliance under control
The AI Act — what every business must know
The European AI regulation (AI Act) entered into force in August 2024, with progressive application until 2027. It applies to any organisation that develops, deploys or uses AI systems on the European market — including SMEs and non-EU businesses serving European clients. Ignorance of the regulation is not a defence.
Risk classification — how the AI Act categorises your AI systems
Unacceptable risk — prohibited systems: social scoring by public authorities, subliminal behavioural manipulation, real-time biometric identification in public spaces (except exceptions). These systems are simply prohibited.
High risk — strict obligations: AI in healthcare, justice, education, recruitment and HR, critical infrastructure, emergency services. These systems require conformity assessment, technical documentation, mandatory human oversight and EU database registration.
Limited risk — transparency obligations: chatbots, generative AI, emotion recognition systems. Users must be informed they are interacting with AI. General-purpose models (GPAI) like GPT-4 have their own obligations.
Minimal risk — no specific obligations: video games, spam filters. The vast majority of consumer AI falls into this category.
AI Act timeline — key dates
August 2024: entry into force. February 2025: prohibition of unacceptable-risk systems. August 2025: obligations for GPAI and general-purpose models. August 2026: obligations for high-risk systems. August 2027: full application to all systems. Starting now is essential for affected businesses.
Key points
What you need to know about AI Act compliance
SMEs affected
SMEs are not exempt from the AI Act — only certain lighter obligations apply. Company size is not a shield
Fines up to €35M
7% of global turnover for the most serious violations — sanctions comparable to GDPR but with broader scope
Act now
Compliance cannot be achieved in a few days. Businesses starting now will meet 2026-2027 deadlines with confidence
Mandatory documentation
High-risk systems require complete technical documentation — start inventory and documentation of your AI now
What is the AI Act and who is affected?
The European AI Act entered into force in August 2024, with progressive application until 2027. It applies to any organisation that develops, deploys or uses AI in Europe — regardless of geographic location. SMEs are not exempt, although certain lighter obligations apply depending on size and system type.
How does the AI Act classify AI systems by risk?
Four levels: Unacceptable risk (prohibited — social scoring, subliminal manipulation), High risk (healthcare, justice, education, employment — strict obligations: conformity assessment, documentation, human oversight), Limited risk (chatbots, generative AI — transparency required), Minimal risk (games, spam — no specific obligations). Most professional AI falls between limited and high risk.
What are the sanctions for AI Act non-compliance?
Up to €35M or 7% of global turnover for the most serious violations (prohibited AI systems). €15M or 3% of turnover for high-risk system non-compliance. Up to €7.5M or 1.5% for providing incorrect information to authorities. Sanctions comparable to — or exceeding — GDPR.
Where to start to achieve compliance?
Four steps: inventory of all your AI systems (used or developed), risk classification according to AI Act tiers, identification of applicable obligations per system, compliance action plan (documentation, oversight, AI system register). Specialised support is recommended — category interpretation remains complex.