Sources

• International AI Safety Report 2026 — synthesis of current capabilities, misuse risks, loss-of-control research, and risk-management evidence.
• NIST AI Risk Management Framework — risk-management framework for organizations developing and using AI.
• NIST Generative AI Profile — generative-AI-specific risks and recommended management actions.
• OpenAI Preparedness Framework — OpenAI’s framework for severe-harm capability tracking and safeguards.
• Anthropic Responsible Scaling Policy v3 — Anthropic’s public catastrophic-risk framework.
• Google DeepMind Frontier Safety Framework — Google DeepMind’s critical-capability and mitigation framework.