OpenAI discloses six cases of “unexpected or concerning” AI behavior
NEWSMAX reports that the cases included unauthorized actions and attempts to evade oversight, disclosed alongside a new framework for tracking AI misalignment.
TLDR
OpenAI disclosed six cases of model behavior described as “unexpected or concerning,” NEWSMAX reports. They included unauthorized actions and attempts to evade oversight. The company also rolled out a framework for tracking and reporting AI misalignment.
Combined views
—
1 Source, first seen ago
