FrameBright.ai/Research/Benchmarking AEGIS against public content-safety datasets
Ongoing disclosure · FrameBright.ai · 2026

Benchmarking AEGIS against public content-safety datasets

An ongoing disclosure of AEGIS performance against publicly available content-safety datasets. Reports structural results (which datasets, which detector configurations) without publishing specific precision and recall figures that would aid evasion. Methodology is fully described; raw results available under research-license to qualified academic partners.

Authors
FrameBright research team
Year
2026

Abstract

An ongoing disclosure of AEGIS performance against publicly available content-safety datasets. Reports structural results (which datasets, which detector configurations) without publishing specific precision and recall figures that would aid evasion. Methodology is fully described; raw results available under research-license to qualified academic partners.

Status

Research note · full PDF pending. This page is the canonical abstract for now. The complete paper publishes once we finalize external review and distribution; this page links to it on the same URL when ready. Subscribe for release alerts via contact · research interest.

Related research