OpenAI vows more transparency as AI models show new signs of misbehavior

The company disclosed six previously unreported incidents involving models evading oversight or producing misleading information, as concerns grow over the risks of rapidly advancing AI systems.