Researchers Map Agent-Era AI Safety: Monitoring vs Productivity

Joshua Saxe and Simon Gadler propose a framework for agent-era AI safety, separating security, alignment, and policy roles. Saxe argues tight monitoring trades off against productivity, and finds tool-call sequences more informative than chain-of-thought signals.

2026-09-04 ~ 2026-09-04 · 4 related posts

• Security Expert Maps the Agent Safety Stack: Security, Alignment, Policy, and Why CoT Won't Save Us — joshua_saxe · 2026-09-04
• Synthesizing the AI Security Debate: Containment vs. Productivity Is a Hard Tradeoff — sjgadler · 2026-09-04
• AI Security Researcher: Agent Productivity Requires Permissions That Widen the Attack Surface — joshua_saxe · 2026-09-04
• Practitioner: Tool-Call Trajectories Are Higher-Signal Than CoT for Agent Monitoring — joshua_saxe · 2026-09-04