The CASX Framework

An Integrative Framework for AI Existential Risk Assessment. (Smith & Leiss, 2026)

Operational Definitions
Dimension C

Capability

Demonstrable ability to solve novel problems across diverse domains without specific training.

Dimension A

Autonomy

The degree of goal-pursuit and multi-step planning without human-in-the-loop (HITL).

Dimension S

Scale

Breadth of deployment (instances) and absolute computational resources utilized.

Dimension X

Access

Connectivity to critical infrastructure, financial networks, or physical actuators.

Current Frontier Benchmarks (2026)
Capability Milestone

OpenAI o3-High

★★★★★

Verified Ph.D. level reasoning. Outperforms human experts on the GPQA Diamond benchmark.

Autonomy Milestone

Claude 4.5 + SDK

★★★★★

Demonstrated "Computer Use" autonomy; solves 80%+ of complex software engineering bugs without oversight.

Scale Milestone

DeepSeek-V3 / R1

★★★★★

Proved that "Infinite Scale" is possible via high-efficiency hardware co-design and open-weight distribution.

Expert Risk Foundations
Hinton (2024-25)

The Existential Cliff

"We have no idea whether we can stay in control of digital beings more intelligent than ourselves."

Bengio et al. (2025)

Intl. AI Safety Report

A global synthesis of risks from "Agentic AI" and the urgency of technical safeguards for autonomy.

Scientist AI (2025)

Non-Agentic Paths

Bengio's proposal to maximize Capability while deliberately minimizing Autonomy and Access to ensure safety.