AI Capabilities
Overview
Section titled “Overview”This section documents the evolving capabilities of AI systems and their implications for safety. Understanding what AI can do—and what it might soon be able to do—is essential for anticipating risks and designing appropriate safeguards.
Capability Domains
Section titled “Capability Domains”Core Capabilities
Section titled “Core Capabilities”- Language Models - Text generation and understanding
- Reasoning - Multi-step logical inference
- Coding - Software development assistance
Emerging Capabilities
Section titled “Emerging Capabilities”- Agentic AI - Autonomous goal-directed behavior
- Long-Horizon Planning - Extended strategic reasoning
- Tool Use - Interfacing with external systems
Safety-Critical Capabilities
Section titled “Safety-Critical Capabilities”- Situational Awareness - Understanding of own context
- Self-Improvement - Recursive capability enhancement
- Persuasion - Influencing human decisions
Applied Capabilities
Section titled “Applied Capabilities”- Scientific Research - Accelerating discovery
- Persuasion - Influence and manipulation potential
Why Capabilities Matter for Safety
Section titled “Why Capabilities Matter for Safety”Capability levels determine:
- Which risks become active - Many risks only emerge at certain capability thresholds
- How much time remains - Faster capability growth compresses safety timelines
- What interventions are viable - Some approaches only work before certain capabilities emerge
Capability Profiles Include
Section titled “Capability Profiles Include”- Current state - What models can do today
- Trajectory - How capability is improving
- Safety implications - What risks this enables
- Measurement approaches - How to evaluate this capability