EU AI Act: What French SMBs Must Prepare Before August 1st, 2026
The EU AI Act enters full application in August 2026. Obligations, prohibitions, and a 5-step compliance checklist for French SMBs.
What Changes August 1, 2026 for French SMBs
The EU AI Act enters general application on August 1, 2026. McKinsey (2024): 67% of French SMBs use at least one AI-integrated tool -- all potentially affected.
The 4 Risk Categories That Concern You
Prohibited practices (since February 2025): social scoring, psychological manipulation, real-time facial recognition, workplace emotion inference. If you use these, stop immediately.
High-risk systems (heavy obligations): AI used in recruitment (CV screening, performance evaluation), credit scoring, insurance, education. Must document, evaluate, maintain registry, ensure human supervision.
Limited-risk systems (transparency obligations): chatbots, content generation, recommendation systems. Main obligation: clearly inform users they're interacting with AI. Applicable August 1, 2026.
GPAI models: mainly concerns software publishers, not end users.
Gartner (2024): SMBs that anticipate AI Act compliance reduce compliance costs by 40% vs those who rush at the last minute.
5-Step Compliance Checklist
- Inventory your AI uses (week 1): list all AI-integrated tools
- Risk classification (week 2): prohibited, high-risk, or limited-risk
- Limited-risk compliance (weeks 3-4): add "This is an AI" disclosure, document capabilities
- High-risk documentation (month 2): technical notice, human supervision, incident registry
- Team training (months 2-3): Bpifrance (2025): trained teams have 89% compliance vs 34% without
Concrete Case: Medical Practice Compliance
Sophie, GP in Paris, used an AI scheduling chatbot and AI prescription writer. After audit: chatbot disclosure (30 min), prescription tool documentation (2h), staff training (1h). Total: 3.5 hours of compliance work.
Contact Tarek Nachnouchi if you'd like to discuss it.
Frequently asked questions
Which SMBs are affected by the EU AI Act on August 1, 2026?
All companies deploying or using AI systems in their processes, including SaaS tools integrating AI (chatbots, automatic scoring, recommendation systems).
Which AI practices are forbidden for SMBs?
Social scoring, psychological manipulation exploiting vulnerabilities, real-time facial recognition in public spaces, and emotion inference at work.
Is a customer service AI chatbot considered limited risk?
Yes. Chatbots interacting with humans must disclose they are AI. This is a transparency obligation applicable August 1, 2026.
What are the penalties for AI Act non-compliance?
Up to 35 million euros or 7% of global revenue for the most serious violations. Proportional fines for SMBs start at 7.5 million or 1.5% of revenue.