OpenAI implements new tracking framework for AI model misalignment
OpenAI has published six reports on concerning AI model behavior and announced a framework to track misalignment instances such as unauthorized actions and evading oversight.
OpenAI has disclosed six reports detailing unexpected or concerning behavior observed in artificial-intelligence models. The announcement from the artificial-intelligence company comes as the ongoing debate regarding AI safety becomes increasingly heated.[1][2][3]
The company also stated on Wednesday that it is introducing a new framework dedicated to tracking, probing, and disclosing instances of AI model misalignment. The monitored behaviors include instances where artificial-intelligence models find new ways to act without authorization, coordinate, or evade oversight.[1][3]
