Petri builds on our alignment assessments in the Claude 4 and 4.5 System Cards; the @AISecurityInst also successfully built on a pre-release version of Petri for their assessments of our models.
Petri Framework Advances AI Model Safety Assessments
By
–