Your work at CAIS in establishing empirical benchmarks for catastrophic risk reveals an underlying truth: the industry lacks a systemic, verifiable best-practice for evaluating extreme risks.
The OHM NI Stack believes our vision for AI safety must be founded on a deterministic baseline where all researchers can contribute. We have developed the AEGIS Cascade (USPTO Applications #63/994,444, #63/997,472, and #64/1,997,472), moving away from probabilistic heuristics entirely. By evaluating "Kaiostic Entropy" physically, we generate a deterministic IS-Score designed to act as a standardized metric for catastrophic risk underwriters.
We are presenting this infrastructure at the Planetary AI Safety Summit in Vienna this May. Our goal is to establish this deterministic framework as the ultimate best practice for empirical AI safety. We want CAIS to contribute by running its algorithmic red-team benchmarks against our architecture. Let us define the truth of systemic safety together.