AI Act: Countdown Begins for Companies Before August 2, 2026

The European AI Act entered into force on August 1, 2024, and applies progressively. Bans on certain AI practices and the obligation of AI literacy have applied since February 2, 2025. Governance rules and obligations for general-purpose AI models have applied since August 2, 2025. The next major step is August 2, 2026, when many provisions, particularly transparency obligations under Article 50, become generally applicable. The Digital Omnibus has delayed high-risk AI system rules, with application now expected on December 2, 2027, for autonomous systems in Annex III, and August 2, 2028, for systems in regulated products, but this does not suspend other parts of the regulation. From August 2, 2026, transparency obligations become central, covering chatbots, AI-generated content, deepfakes, and synthetic media. The European Commission published a Code of Good Practices on June 10, 2026, to help providers and deployers comply with Article 50 marking and detection obligations. Companies face concrete questions: whether clients know they interact with a chatbot, whether public-interest content is flagged as AI-generated, and whether synthetic images or voices are labeled. The biggest risk is invisible AI: AI features in CRMs, marketing scoring modules, HR assistants, content generators, and office copilots often lack mapping. Shadow AI, informal use of general tools, and AI in third-party deliverables are common. The Hub France IA recommends building an inventory of AI systems and maintaining a registry with data sources, technical architectures, business use cases, and risk analysis. Compliance starts with a simple registry of AI uses, identifying tools, services, purposes, data, affected persons, and roles. A pragmatic checklist includes appointing an AI Act lead, inventorying AI uses, seeking invisible AI, creating a minimal registry, identifying sensitive uses (HR, recruitment, credit, health), preparing transparency measures, reviewing supplier contracts, training teams, implementing supervision, and building a proof dossier. Sanctions are significant: up to 35 million euros or 7% of global turnover for prohibited practices, 15 million or 3% for other breaches including transparency, and 7.5 million or 1% for incorrect information. However, immediate risks include losing tenders, failing supplier questionnaires, or not being able to prove human supervision. Companies should not wait for perfect compliance; they should start with a first mapping and registry now.