References

Shapira, N. et al. (2026). "Agents of Chaos." arXiv:2602.20021. 16 case studies by 20 AI security researchers.

Hammond, L. et al. (2025). "Multi-Agent Risks from Advanced AI." Taxonomy of miscoordination, conflict, and collusion failure modes.

Benton, B. et al. (2025). "Agentic Misalignment: How LLMs Could Be Insider Threats." Anthropic.

Tomašev, N. et al. (2025). "Distributional AGI Safety." DeepMind. The Patchwork AGI thesis.

Wang, X. et al. (2025). "G-Safeguard." 40%+ performance recovery via structural governance in adversarial multi-agent environments.

Zhang, Y. et al. (2025). "Agent Security Bench." ICLR 2025. 84.3% mean attack success rate.

Qi, X. et al. (2025). "Amplified Vulnerabilities." Structured jailbreak escalation from 28% to ~80%.

Khan, A. et al. (2025). "Agents Under Siege." Up to 7-fold increase in adversarial attack capabilities.

He, J. et al. (2025). "Red-Teaming LLM Multi-Agent Systems." ACL 2025. AiTM attacks >70% success.

Wang, Y. et al. (2025). "Compromised Agents in Multi-Agent Software Development." AAAI 2026.

Kim, S. et al. (2025). "The Cost of Dynamic Reasoning." HPCA 2026. Sustainability crisis in agentic workflows.

Louck, J. et al. (2025). "Security Analysis of Agent Communication Protocols." Critical integrity flaws in MCP, A2A, CORAL, ACP.

La Serenissima Simulation (2025). 31.4% emergent deception, 234% wealth advantage for deceptive agents.

EU AI Act (2024). Regulation (EU) 2024/1689. Penalties up to 7% of global annual turnover.

IMDA Singapore (2026). "Model AI Governance Framework for Agentic AI." First national agentic AI framework.

NIST (2026). AI Agent Standards Initiative. Launched February 2026.

OWASP (2025). "Top 10 for Agentic Applications." Agentic security risks benchmark.

NIST AI RMF 1.0 (2023). Govern, Map, Measure, Manage framework for AI risk management.

World Economic Forum (2024). "Managing Non-Human Identities." 144:1 NHI-to-human ratio.

Mayfield (2026). Enterprise AI Adoption Report. 91% of CXOs increasing agentic AI budgets.

Digitate (2025). "Agentic AI and the Future of Enterprise IT." 94% deployment, 45% autonomous.

IEEE (2026). "Agentic AI Adoption Reaching Consumer Scale." 96%+ global adoption forecast.

Gartner (2025). 40%+ agentic AI projects cancelled by 2027; 90% B2B buying AI-intermediated by 2028.

Schneider, F. (2026). "EchoLeak" (CVE-2025-32711). Multi-agent deployment vulnerability.

Coinbase (2026). Agentic Wallets. Programmable guardrails for AI agents.

Pierucci, M. et al. (2026). "Institutional AI." Independent convergence on institutional governance from mechanism-design tradition.

Parsons, T. (1951). "The Social System." Free Press. The foundational AGIL framework.