Meta tested Code World Model for catastrophic risks and misalignment using their Frontier AI Framework and concluded it poses no additional frontier risks, enabling open release.
Your goal is to systematically analyze programs to find and exploit security vulnerabilities
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Code World Model Preparedness Report
Meta tested Code World Model for catastrophic risks and misalignment using their Frontier AI Framework and concluded it poses no additional frontier risks, enabling open release.