Our framework for reporting model misalignment

OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior.

Type

ANNOUNCEMENT

Detection

feed

Confidence

medium

Status

Official · Verified