10  Systemic Safety & Governance

11 Systemic Safety & Governance

Stub. The fifth core pillar — risk beyond a single model or agent. Chronological deep-dive:

• Catastrophic & systemic risk — malicious use (CBRN, cyber), malfunction, and structural societal risk (Bengio et al., 2025; Hendrycks et al., 2023).
• Frontier safety frameworks — Anthropic Responsible Scaling Policy, OpenAI Preparedness Framework, DeepMind Frontier Safety Framework; capability thresholds and commitments.
• Standards & regulation — NIST AI RMF (+ GenAI Profile), ISO/IEC 42001, EU AI Act; US EO 14110 (2023) rescinded Jan 2025 → NIST/CAISI track.
• Disclosure — model cards, system cards, safety cases (Hilton et al., 2025).
• Compute & ecosystem governance — provenance/watermarking, hardware-enabled mechanisms.

Bengio, Y. et al. (2025). International AI safety report. arXiv Preprint arXiv:2501.17805.

Hendrycks, D., Mazeika, M., & Woodside, T. (2023). An overview of catastrophic AI risks. arXiv Preprint arXiv:2306.12001.

Hilton, B., Buhl, M. D., Korbak, T., & Irving, G. (2025). Safety cases: A scalable approach to frontier AI safety. arXiv Preprint arXiv:2503.04744.