04:00
2026-05-28
arxiv.org
ai-safety
Intelligence as Managed Autonomy: Failure, Escalation, and Governance for Agentic AI Systems
A new paper from researchers proposes the SMARt framework, a four-layer model designed to manage failures in autonomous AI systems by formally detecting epistemic drift and mandating the suspension or…