Guidelight Releases AI Transparency Standard Requiring Public Risk Assessments

sjgadler · x · 2026-08-08

Guidelight has released the Transparency Standard (v1.0) for AI developers. The standard aims to make safety-relevant information legible to stakeholders outside of AI labs, including scientists, civil society, and governments.

The core principle requires developers to expose their structured risk assessments of relevant models to public scrutiny. The assessments must address models capable of catastrophic risks (e.g., CBRN weapons, autonomous cyberattacks), allowing developers to set their own severity thresholds (e.g., more than 100 deaths or $1 billion in property damage).

Original post →

More from Safety

Safety channel →