Twelve frontier AI safety frameworks score between 8% and 34% on adapted risk-management criteria, with a median of 18%, leaving them too vague to serve as reliable external accountability mechanisms.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evaluating AI Providers' Frontier Safety Frameworks
Twelve frontier AI safety frameworks score between 8% and 34% on adapted risk-management criteria, with a median of 18%, leaving them too vague to serve as reliable external accountability mechanisms.