EU AI Act Enforcement Is Here: Chatbot Rules Live, High-Risk AI Delay Now Binding Law

What happened

On July 10, 2026, the EU activated another enforcement phase of the EU AI Act: transparency requirements for AI systems interacting with humans (chatbots) are now in effect.

Context and impact

EU AI Act enforcement has been rolling out in waves since 2025. The new chatbot rules affect all major AI platforms available in the EU — ChatGPT, Claude, Gemini, Perplexity. Companies risk fines up to 3% of global turnover for violations. Grok 4.5 is a concrete example: xAI did not release it in the EU because the model has not yet completed EU AI Act audit requirements for systemic-risk models.

Details

• Chatbot transparency rules (Article 50): obligation to inform users they are interacting with AI — effective July 10, 2026
• Systemic risk: frontier models above 10^25 FLOP — mandatory adversarial testing, audit, incident reporting
• Grok 4.5 EU gap: xAI targets mid-July for EU access after completing audits
• Next phase: rules for high-risk AI systems — August 2027
• EU also released a coordinated Cybersecurity & AI Action Plan for member states
• Fines for violations: up to 3% of global turnover or EUR 15 million