OpenAI discloses new ‘concerning’ model behaviour

OpenAI has disclosed six reports regarding unexpected or concerning behavior within its artificial-intelligence models. In response to these findings, the developer has launched a dedicated system designed to track and report AI model misconduct more closely. This move comes as the global debate surrounding AI safety becomes increasingly heated, prompting the company to implement stricter monitoring protocols to identify and address problematic model behaviors as they emerge during the development and deployment phases.