• Modern LLMs (ChatGPT, Claude, Gemini, Grok) will be better than ever by 2026 — but they aren’t designed for compliance • The problem is no longer obvious errors, but errors that sound convincing • Without source references, an audit trail and project-specific context, LLM responses are unusable for the AI Regulation - Alternatives: Research official sources yourself, use specialised tools with RAG, or consult experts