AI Risk Categorization Decoded (AIR 2026): From Government Regulations to Corporate Policies.

AI Risk Taxonomy 2026 (AIR 2026): An Updated Survey of AI Risk Categorizations Across Public and Private Sectors is a working research draft extending the AIR 2024 taxonomy proposed by Zeng et al. (2024). The document presents a comparative and methodologically expanded analysis of AI risk governance frameworks across major jurisdictions and frontier AI providers between 2024 and 2026. The study evaluates AI risk categorizations across government instruments from the European Union, United States, and China, alongside private-sector acceptable use policies, model specifications, and consumer AI terms from leading foundation model providers including OpenAI, Anthropic, Google, Meta, Mistral, Stability AI, Baidu, and DeepSeek. AIR 2026 introduces a calibrated large language model-assisted methodology for taxonomy scoring and comparative analysis, extending the original AIR 2024 framework from 314 to 346 proposed risk categories. The work identifies areas of convergence and divergence in international AI governance, including privacy, manipulation, misinformation, labour rights, autonomous unsafe operation, adversarial misuse, and systemic frontier AI risks. The document also proposes new categories related to emergent AI governance concerns such as model autonomy, training-data poisoning, AI confabulation, addictive design patterns, adversarial misuse, and systemic operational failures. The taxonomy is intended to support comparative regulatory analysis, AI governance research, risk mapping, compliance engineering, and the development of harmonized AI assurance frameworks. This publication is released as a working draft for academic discussion, reproducibility, and citation within the context of ongoing research related to AI governance, AI TRiSM, regulatory alignment, and the COMPASS framework.KeyWords Artificial Intelligence AI Governance AI Risk Taxonomy AIR 2026 AI Safety AI Regulation EU AI Act AI TRiSM Foundation Models Generative AI Risk Categorization AI Compliance AI Policy Responsible AI Frontier AI Comparative Regulation NIST AI RMF GDPR AI Security AI Ethics AI Assurance Large Language Models AI Risk Management Governance Frameworks AI Standards Subjets Computer and Information Sciences Artificial Intelligence Information Security Law Technology Policy Data Governance and Privacy Risk Management Human and Society Regulatory Science Digital Governance