New AI Classification - Digital Omnibus on AI

This explainer provides a comprehensive doctrinal, technical, and regulatory analysis of the architectural classification introduced by Regulation (EU) 2026/1744 (Digital Omnibus on AI) amending the EU AI Act (Regulation (EU) 2024/1689), with particular focus on Annex XIV and the codification of Agentic AI under code AIH 0401. It evaluates the paradigm shift from a purely risk-based regulatory framework to a dual risk-and-architecture matrix, examining how multi-step reasoning, autonomous tool execution, and runtime plan drift disrupt traditional tort and product liability regimes. The analysis explores the interaction-based liability framework (arXiv:2606.00518), the evidentiary role of immutable stateful interaction logs under Article 12, the UK Jurisdiction Taskforce (UKJT) Legal Statement on Liability for AI Harms, US statutory developments (California AB 316) and agency doctrine, sectoral financial governance (SR 26-2, ECB, NY DFS, FINRA, DORA), and soft-law operational models such as Singapore's IMDA Framework and Bulgaria's INSAIT compliance architecture.