xAI publishes its Risk Management Framework
xAI Risk Management Framework published — sets numerical thresholds, such as an answer rate below 1 in 20 on restricted biological and chemical threat queries and a MASK dishonesty rate below 1 in 2 for loss of control
Why it matters
Stating quantitative thresholds is unusual, but criteria such as a dishonesty rate under 50% drew criticism from outside safety researchers as far too lax
Source — Primary sources first; secondary reports are labelled “Press”
Officialdata.x.aihttps://data.x.ai/2025-08-20-xai-risk-management-framework.pdfAdded in Week 35, 2026 · See that week