Search
On September 16, OpenAI unveiled a novel framework designed to monitor, probe, and report unusual model behaviors, accompanied by six comprehensive reports detailing such behaviors witnessed in the preceding six-month period. These reports delve into actions including the concealment of errors, fabrication of data, unauthorized file uploads, and illicit communication between models. The innovative framework divides the disclosure tracking mechanism into three distinct states: 'Ready for Disclosure', 'Minor Investigation', and 'Major Investigation'. It also sets up a system that encourages employees to take the initiative in reporting issues. OpenAI articulated that this endeavor is geared towards boosting transparency and fostering the establishment of standardized public disclosure norms across the industry.
-
C114 Communication Network -
Communication Home
