Longterm Wiki

Tool-Use Restrictions

Tool-use restrictions limit what actions and APIs AI systems can access, directly constraining their potential for harm. This approach is critical for agentic AI systems, providing hard limits on capabilities regardless of model intentions, with METR evaluations showing agentic task completion horizons doubling every 7 months.

Related

Related Pages

Top Related Pages

Analysis

Goal Misgeneralization Probability Model

Approaches

Structured Access / API-OnlyMulti-Agent SafetyCircuit Breakers / Inference InterventionsCapability ElicitationThird-Party Model AuditingDangerous Capability Evaluations

Concepts

Alignment Deployment OverviewTool Use and Computer Use

Tags

agent-safetycapability-restrictionsdefense-in-depthdeployment-safetypermission-systemsmcp-security